线性回归算法.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "使用数据集预测房价"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Features\n",
    "\n",
    "<table>\n",
    "  <tbody>    \n",
    "    <tr valign=\"top\">\n",
    "      <td valign=\"top\">\n",
    "        <ul>\n",
    "          <li>MoSold: Month Sold</li>\n",
    "          <li>YrSold: Year Sold</li><br>\n",
    "          \n",
    "          <li>SaleType: Type of sale</li>\n",
    "          <li>SaleCondition: Condition of sale</li><br>\n",
    "\n",
    "          <li>MSSubClass: The building class</li>\n",
    "          <li>MSZoning: The general zoning classification</li><br>\n",
    "\n",
    "          <li>Neighborhood: Physical locations within Ames city limits</li>\n",
    "          <li>Street: Type of road access</li>\n",
    "          <li>Alley: Type of alley access</li><br>\n",
    "\n",
    "          <li>LotArea: Lot size in square feet</li>\n",
    "          <li>LotConfig: Lot configuration</li>\n",
    "          <li>LotFrontage: Linear feet of street connected to property</li>\n",
    "          <li>LotShape: General shape of property</li><br>\n",
    "\n",
    "          <li>LandSlope: Slope of property</li>\n",
    "          <li>LandContour: Flatness of the property</li><br>\n",
    "\n",
    "          <li>YearBuilt: Original construction date</li>\n",
    "          <li>YearRemodAdd: Remodel date</li>\n",
    "          <li>OverallQual: Overall material and finish quality</li>\n",
    "          <li>OverallCond: Overall condition rating</li><br>\n",
    "\n",
    "          <li>Utilities: Type of utilities available</li>\n",
    "          <li>Foundation: Type of foundation</li>\n",
    "          <li>Functional: Home functionality rating</li><br>\n",
    "\n",
    "          <li>BldgType: Type of dwelling</li>\n",
    "          <li>HouseStyle: Style of dwelling</li><br>\n",
    "          \n",
    "          <li>1stFlrSF: First Floor square feet</li>\n",
    "          <li>2ndFlrSF: Second floor square feet</li>\n",
    "          <li>LowQualFinSF: Low quality finished square feet (all floors)</li>\n",
    "          <li>GrLivArea: Above grade (ground) living area square feet</li>\n",
    "          <li>TotRmsAbvGrd: Total rooms above grade (does not include bathrooms)</li><br>\n",
    "\n",
    "          <li>Condition1: Proximity to main road or railroad</li>\n",
    "          <li>Condition2: Proximity to main road or railroad (if a second is present)</li><br>\n",
    "          \n",
    "          <li>RoofStyle: Type of roof</li>\n",
    "          <li>RoofMatl: Roof material</li><br>\n",
    "\n",
    "          <li>ExterQual: Exterior material quality</li>\n",
    "          <li>ExterCond: Present condition of the material on the exterior</li>\n",
    "          <li>Exterior1st: Exterior covering on house</li>\n",
    "          <li>Exterior2nd: Exterior covering on house (if more than one material)</li><br><br>\n",
    "          \n",
    "        </ul>\n",
    "      </td>\n",
    "      \n",
    "      <td valign=\"top\">\n",
    "        <ul>\n",
    "          <li>MasVnrType: Masonry veneer type</li>\n",
    "          <li>MasVnrArea: Masonry veneer area in square feet</li><br>\n",
    "          \n",
    "          <li>WoodDeckSF: Wood deck area in square feet</li>\n",
    "          <li>OpenPorchSF: Open porch area in square feet</li>\n",
    "          <li>EnclosedPorch: Enclosed porch area in square feet</li>\n",
    "          <li>3SsnPorch: Three season porch area in square feet</li>\n",
    "          <li>ScreenPorch: Screen porch area in square feet</li><br>\n",
    "\n",
    "          <li>PoolArea: Pool area in square feet</li>\n",
    "          <li>PoolQC: Pool quality</li>\n",
    "          <li>Fence: Fence quality</li>\n",
    "          <li>PavedDrive: Paved driveway</li><br>\n",
    "\n",
    "          <li>GarageType: Garage location</li>\n",
    "          <li>GarageYrBlt: Year garage was built</li>\n",
    "          <li>GarageFinish: Interior finish of the garage</li>\n",
    "          <li>GarageCars: Size of garage in car capacity</li>\n",
    "          <li>GarageArea: Size of garage in square feet</li>\n",
    "          <li>GarageQual: Garage quality</li>\n",
    "          <li>GarageCond: Garage condition</li><br>\n",
    "\n",
    "          <li>Heating: Type of heating</li>\n",
    "          <li>HeatingQC: Heating quality and condition</li>\n",
    "          <li>CentralAir: Central air conditioning</li>\n",
    "          <li>Electrical: Electrical system</li><br>\n",
    "          \n",
    "          <li>FullBath: Full bathrooms above grade</li>\n",
    "          <li>HalfBath: Half baths above grade</li><br>\n",
    "          \n",
    "          <li>BedroomAbvGr: Number of bedrooms above basement level</li><br>\n",
    "          \n",
    "          <li>KitchenAbvGr: Number of kitchens</li>\n",
    "          <li>KitchenQual: Kitchen quality</li><br>\n",
    "          \n",
    "          <li>Fireplaces: Number of fireplaces</li>\n",
    "          <li>FireplaceQu: Fireplace quality</li><br>\n",
    "          \n",
    "          <li>MiscFeature: Miscellaneous feature not covered in other categories</li>\n",
    "          <li>MiscVal: Value of miscellaneous feature</li><br>\n",
    "          \n",
    "          <li>BsmtQual: Height of the basement</li>\n",
    "          <li>BsmtCond: General condition of the basement</li>\n",
    "          <li>BsmtExposure: Walkout or garden level basement walls</li>\n",
    "          <li>BsmtFinType1: Quality of basement finished area</li>\n",
    "          <li>BsmtFinSF1: Type 1 finished square feet</li>\n",
    "          <li>BsmtFinType2: Quality of second finished area (if present)</li>\n",
    "          <li>BsmtFinSF2: Type 2 finished square feet</li>\n",
    "          <li>BsmtUnfSF: Unfinished square feet of basement area</li>\n",
    "          <li>BsmtFullBath: Basement full bathrooms</li>\n",
    "          <li>BsmtHalfBath: Basement half bathrooms</li>\n",
    "          <li>TotalBsmtSF: Total square feet of basement area</li>\n",
    "        </ul>\n",
    "      </td>\n",
    "    </tr>\n",
    "    \n",
    "  </tbody>\n",
    "</table>\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "data_path = ['dataset']\n",
    "filepath = os.sep.join(data_path + ['Ames_Housing_Sales.csv'])\n",
    "data = pd.read_csv(filepath)#, sep=','"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1379, 80)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.shape#数据集大小"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "object     43\n",
       "float64    21\n",
       "int64      16\n",
       "dtype: int64"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.dtypes.value_counts()#列的类型"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "无序分类应该是one-hot编码，但是这样显着增加特征的数量并创建彼此高度相关的特征，one-hot 编码列的总数是 n-1，其中 n 是类别的数量。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:2: DeprecationWarning: `np.object` is a deprecated alias for the builtin `object`. To silence this warning, use `object` by itself. Doing this will not modify any behavior and is safe. \n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "#选择类型是boject的列\n",
    "mask = data.dtypes == np.object\n",
    "categorical_cols = data.columns[mask]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Alley', 'BldgType', 'BsmtCond', 'BsmtExposure', 'BsmtFinType1',\n",
       "       'BsmtFinType2', 'BsmtQual', 'CentralAir', 'Condition1', 'Condition2',\n",
       "       'Electrical', 'ExterCond', 'ExterQual', 'Exterior1st', 'Exterior2nd',\n",
       "       'Fence', 'FireplaceQu', 'Foundation', 'Functional', 'GarageCond',\n",
       "       'GarageFinish', 'GarageQual', 'GarageType', 'Heating', 'HeatingQC',\n",
       "       'HouseStyle', 'KitchenQual', 'LandContour', 'LandSlope', 'LotConfig',\n",
       "       'LotShape', 'MSZoning', 'MasVnrType', 'MiscFeature', 'Neighborhood',\n",
       "       'PavedDrive', 'PoolQC', 'RoofMatl', 'RoofStyle', 'SaleCondition',\n",
       "       'SaleType', 'Street', 'Utilities'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "categorical_cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Neighborhood     25\n",
       "Exterior2nd      16\n",
       "Exterior1st      14\n",
       "SaleType          9\n",
       "Condition1        9\n",
       "HouseStyle        8\n",
       "RoofMatl          8\n",
       "Condition2        8\n",
       "Functional        7\n",
       "BsmtFinType2      7\n",
       "FireplaceQu       6\n",
       "Foundation        6\n",
       "RoofStyle         6\n",
       "SaleCondition     6\n",
       "BsmtFinType1      6\n",
       "Heating           6\n",
       "GarageType        6\n",
       "BsmtExposure      5\n",
       "Electrical        5\n",
       "Fence             5\n",
       "GarageCond        5\n",
       "GarageQual        5\n",
       "BsmtQual          5\n",
       "LotConfig         5\n",
       "MSZoning          5\n",
       "MiscFeature       5\n",
       "BldgType          5\n",
       "HeatingQC         5\n",
       "KitchenQual       4\n",
       "LandContour       4\n",
       "BsmtCond          4\n",
       "LotShape          4\n",
       "MasVnrType        4\n",
       "ExterQual         4\n",
       "ExterCond         4\n",
       "PoolQC            4\n",
       "Alley             3\n",
       "GarageFinish      3\n",
       "LandSlope         3\n",
       "PavedDrive        3\n",
       "CentralAir        2\n",
       "Street            2\n",
       "Utilities         2\n",
       "dtype: int64"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#查看会创建多少额外的列\n",
    "num_ohc_cols = (data[categorical_cols]\n",
    "                .apply(lambda x: x.nunique())\n",
    "                .sort_values(ascending=False))\n",
    "num_ohc_cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "215"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 如果只有一个值则不需要编码\n",
    "small_num_ohc_cols = num_ohc_cols.loc[num_ohc_cols>1]\n",
    "#one hot编码比类别少一\n",
    "small_num_ohc_cols -= 1\n",
    "#\n",
    "small_num_ohc_cols.sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- .copy() 方法为独热编码创建一个完全独立的数据帧副本\n",
    "- 独热编码后一定要删除原始列\n",
    "- 不是one-hot 编码的数据，需要删除string类型的离散值\n",
    "\n",
    "使用LabelEncoder或者 DictVectorizer将string类型的离散值数值化"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:13: DeprecationWarning: `np.int` is a deprecated alias for the builtin `int`. To silence this warning, use `int` by itself. Doing this will not modify any behavior and is safe. When replacing `np.int`, you may wish to use e.g. `np.int64` or `np.int32` to specify the precision. If you wish to review your current use, check the release note link for additional information.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  del sys.path[0]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import OneHotEncoder, LabelEncoder\n",
    "\n",
    "# 复制数据\n",
    "data_ohc = data.copy()\n",
    "\n",
    "# 编码器\n",
    "le = LabelEncoder()\n",
    "ohc = OneHotEncoder()\n",
    "\n",
    "for col in num_ohc_cols.index:\n",
    "    \n",
    "    # 整数编码字符串类别\n",
    "    dat = le.fit_transform(data_ohc[col]).astype(np.int)\n",
    "     \n",
    "    # 删除原始列\n",
    "    data_ohc = data_ohc.drop(col, axis=1)\n",
    "\n",
    "    # One hot 编码数据--返回稀疏array\n",
    "    new_dat = ohc.fit_transform(dat.reshape(-1,1))\n",
    "   \n",
    "    # 创建不一样的列名\n",
    "    n_cols = new_dat.shape[1]\n",
    "    col_names = ['_'.join([col, str(x)]) for x in range(n_cols)]\n",
    "\n",
    "    # 创建新的 dataframe\n",
    "    new_df = pd.DataFrame(new_dat.toarray(), \n",
    "                          index=data_ohc.index, \n",
    "                          columns=col_names)\n",
    "\n",
    "    # Append the new data to the dataframe\n",
    "    data_ohc = pd.concat([data_ohc, new_df], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "215"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#列数量的差异\n",
    "data_ohc.shape[1] - data.shape[1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "80\n",
      "37\n"
     ]
    }
   ],
   "source": [
    "print(data.shape[1])\n",
    "\n",
    "# 删除data的string列\n",
    "data = data.drop(num_ohc_cols.index, axis=1)\n",
    "\n",
    "print(data.shape[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>1stFlrSF</th>\n",
       "      <th>2ndFlrSF</th>\n",
       "      <th>3SsnPorch</th>\n",
       "      <th>BedroomAbvGr</th>\n",
       "      <th>BsmtFinSF1</th>\n",
       "      <th>BsmtFinSF2</th>\n",
       "      <th>BsmtFullBath</th>\n",
       "      <th>BsmtHalfBath</th>\n",
       "      <th>BsmtUnfSF</th>\n",
       "      <th>EnclosedPorch</th>\n",
       "      <th>...</th>\n",
       "      <th>OverallQual</th>\n",
       "      <th>PoolArea</th>\n",
       "      <th>ScreenPorch</th>\n",
       "      <th>TotRmsAbvGrd</th>\n",
       "      <th>TotalBsmtSF</th>\n",
       "      <th>WoodDeckSF</th>\n",
       "      <th>YearBuilt</th>\n",
       "      <th>YearRemodAdd</th>\n",
       "      <th>YrSold</th>\n",
       "      <th>SalePrice</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>856.0</td>\n",
       "      <td>854.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>706.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>150.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>8</td>\n",
       "      <td>856.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2003</td>\n",
       "      <td>2003</td>\n",
       "      <td>2008</td>\n",
       "      <td>208500.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1262.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>978.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>284.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6</td>\n",
       "      <td>1262.0</td>\n",
       "      <td>298.0</td>\n",
       "      <td>1976</td>\n",
       "      <td>1976</td>\n",
       "      <td>2007</td>\n",
       "      <td>181500.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>920.0</td>\n",
       "      <td>866.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>486.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>434.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6</td>\n",
       "      <td>920.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2001</td>\n",
       "      <td>2002</td>\n",
       "      <td>2008</td>\n",
       "      <td>223500.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>961.0</td>\n",
       "      <td>756.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>216.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>540.0</td>\n",
       "      <td>272.0</td>\n",
       "      <td>...</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7</td>\n",
       "      <td>756.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1915</td>\n",
       "      <td>1970</td>\n",
       "      <td>2006</td>\n",
       "      <td>140000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1145.0</td>\n",
       "      <td>1053.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4</td>\n",
       "      <td>655.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>490.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>8</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9</td>\n",
       "      <td>1145.0</td>\n",
       "      <td>192.0</td>\n",
       "      <td>2000</td>\n",
       "      <td>2000</td>\n",
       "      <td>2008</td>\n",
       "      <td>250000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1374</th>\n",
       "      <td>953.0</td>\n",
       "      <td>694.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>953.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7</td>\n",
       "      <td>953.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1999</td>\n",
       "      <td>2000</td>\n",
       "      <td>2007</td>\n",
       "      <td>175000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1375</th>\n",
       "      <td>2073.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>790.0</td>\n",
       "      <td>163.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>589.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>6</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>7</td>\n",
       "      <td>1542.0</td>\n",
       "      <td>349.0</td>\n",
       "      <td>1978</td>\n",
       "      <td>1988</td>\n",
       "      <td>2010</td>\n",
       "      <td>210000.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1376</th>\n",
       "      <td>1188.0</td>\n",
       "      <td>1152.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>4</td>\n",
       "      <td>275.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>877.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>7</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>9</td>\n",
       "      <td>1152.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>1941</td>\n",
       "      <td>2006</td>\n",
       "      <td>2010</td>\n",
       "      <td>266500.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1377</th>\n",
       "      <td>1078.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>2</td>\n",
       "      <td>49.0</td>\n",
       "      <td>1029.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>112.0</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>5</td>\n",
       "      <td>1078.0</td>\n",
       "      <td>366.0</td>\n",
       "      <td>1950</td>\n",
       "      <td>1996</td>\n",
       "      <td>2010</td>\n",
       "      <td>142125.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1378</th>\n",
       "      <td>1256.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>3</td>\n",
       "      <td>830.0</td>\n",
       "      <td>290.0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>136.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>...</td>\n",
       "      <td>5</td>\n",
       "      <td>0.0</td>\n",
       "      <td>0.0</td>\n",
       "      <td>6</td>\n",
       "      <td>1256.0</td>\n",
       "      <td>736.0</td>\n",
       "      <td>1965</td>\n",
       "      <td>1965</td>\n",
       "      <td>2008</td>\n",
       "      <td>147500.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1379 rows × 37 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "      1stFlrSF  2ndFlrSF  3SsnPorch  BedroomAbvGr  BsmtFinSF1  BsmtFinSF2  \\\n",
       "0        856.0     854.0        0.0             3       706.0         0.0   \n",
       "1       1262.0       0.0        0.0             3       978.0         0.0   \n",
       "2        920.0     866.0        0.0             3       486.0         0.0   \n",
       "3        961.0     756.0        0.0             3       216.0         0.0   \n",
       "4       1145.0    1053.0        0.0             4       655.0         0.0   \n",
       "...        ...       ...        ...           ...         ...         ...   \n",
       "1374     953.0     694.0        0.0             3         0.0         0.0   \n",
       "1375    2073.0       0.0        0.0             3       790.0       163.0   \n",
       "1376    1188.0    1152.0        0.0             4       275.0         0.0   \n",
       "1377    1078.0       0.0        0.0             2        49.0      1029.0   \n",
       "1378    1256.0       0.0        0.0             3       830.0       290.0   \n",
       "\n",
       "      BsmtFullBath  BsmtHalfBath  BsmtUnfSF  EnclosedPorch  ...  OverallQual  \\\n",
       "0                1             0      150.0            0.0  ...            7   \n",
       "1                0             1      284.0            0.0  ...            6   \n",
       "2                1             0      434.0            0.0  ...            7   \n",
       "3                1             0      540.0          272.0  ...            7   \n",
       "4                1             0      490.0            0.0  ...            8   \n",
       "...            ...           ...        ...            ...  ...          ...   \n",
       "1374             0             0      953.0            0.0  ...            6   \n",
       "1375             1             0      589.0            0.0  ...            6   \n",
       "1376             0             0      877.0            0.0  ...            7   \n",
       "1377             1             0        0.0          112.0  ...            5   \n",
       "1378             1             0      136.0            0.0  ...            5   \n",
       "\n",
       "      PoolArea  ScreenPorch  TotRmsAbvGrd  TotalBsmtSF  WoodDeckSF  YearBuilt  \\\n",
       "0          0.0          0.0             8        856.0         0.0       2003   \n",
       "1          0.0          0.0             6       1262.0       298.0       1976   \n",
       "2          0.0          0.0             6        920.0         0.0       2001   \n",
       "3          0.0          0.0             7        756.0         0.0       1915   \n",
       "4          0.0          0.0             9       1145.0       192.0       2000   \n",
       "...        ...          ...           ...          ...         ...        ...   \n",
       "1374       0.0          0.0             7        953.0         0.0       1999   \n",
       "1375       0.0          0.0             7       1542.0       349.0       1978   \n",
       "1376       0.0          0.0             9       1152.0         0.0       1941   \n",
       "1377       0.0          0.0             5       1078.0       366.0       1950   \n",
       "1378       0.0          0.0             6       1256.0       736.0       1965   \n",
       "\n",
       "      YearRemodAdd  YrSold  SalePrice  \n",
       "0             2003    2008   208500.0  \n",
       "1             1976    2007   181500.0  \n",
       "2             2002    2008   223500.0  \n",
       "3             1970    2006   140000.0  \n",
       "4             2000    2008   250000.0  \n",
       "...            ...     ...        ...  \n",
       "1374          2000    2007   175000.0  \n",
       "1375          1988    2010   210000.0  \n",
       "1376          2006    2010   266500.0  \n",
       "1377          1996    2010   142125.0  \n",
       "1378          1965    2008   147500.0  \n",
       "\n",
       "[1379 rows x 37 columns]"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- 创建两个数据集的训练和测试分割。\n",
    "- 为确保数据以相同的方式拆分，请在两个拆分数据集方法中使用相同的 random_state。\n",
    "- 对于每个数据集，在训练数据上拟合一个基本的线性回归模型。\n",
    "- 计算各自模型的训练集和测试集的均方误差。\n",
    " "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "y_col = 'SalePrice'\n",
    "\n",
    "# 拆分不是独热编码的数据集\n",
    "feature_cols = [x for x in data.columns if x != y_col]\n",
    "X_data = data[feature_cols]\n",
    "y_data = data[y_col]\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X_data, y_data, \n",
    "                                                    test_size=0.3, random_state=42)\n",
    "# 拆分是独热编码的数据集\n",
    "feature_cols = [x for x in data_ohc.columns if x != y_col]\n",
    "X_data_ohc = data_ohc[feature_cols]\n",
    "y_data_ohc = data_ohc[y_col]\n",
    "\n",
    "X_train_ohc, X_test_ohc, y_train_ohc, y_test_ohc = train_test_split(X_data_ohc, y_data_ohc, \n",
    "                                                    test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 比较索引确保相等\n",
    "(X_train_ohc.index == X_train.index).all()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>no enc</th>\n",
       "      <th>one-hot enc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>train</th>\n",
       "      <td>1.131507e+09</td>\n",
       "      <td>3.177267e+08</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>test</th>\n",
       "      <td>1.372182e+09</td>\n",
       "      <td>6.048710e+12</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             no enc   one-hot enc\n",
       "train  1.131507e+09  3.177267e+08\n",
       "test   1.372182e+09  6.048710e+12"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "from sklearn.metrics import mean_squared_error\n",
    "\n",
    "LR = LinearRegression()#模型\n",
    "error_df = list()#保存error值的列表\n",
    "#非独热编码\n",
    "LR = LR.fit(X_train, y_train)\n",
    "y_train_pred = LR.predict(X_train)\n",
    "y_test_pred = LR.predict(X_test)\n",
    "\n",
    "error_df.append(pd.Series({'train': mean_squared_error(y_train, y_train_pred),\n",
    "                           'test' : mean_squared_error(y_test,  y_test_pred)},\n",
    "                           name='no enc'))\n",
    " \n",
    "#独热编码\n",
    "LR = LR.fit(X_train_ohc, y_train_ohc)\n",
    "y_train_ohc_pred = LR.predict(X_train_ohc)\n",
    "y_test_ohc_pred = LR.predict(X_test_ohc)\n",
    "\n",
    "error_df.append(pd.Series({'train': mean_squared_error(y_train_ohc, y_train_ohc_pred),\n",
    "                           'test' : mean_squared_error(y_test_ohc,  y_test_ohc_pred)},\n",
    "                          name='one-hot enc'))\n",
    "#连接结果\n",
    "error_df = pd.concat(error_df, axis=1)\n",
    "error_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "请注意，对于训练和测试数据，one-hot 编码数据上的错误值非常不同。\n",
    "尤其是测试数据上的误差要高得多。\n",
    "这是因为 one-hot 编码模型过度拟合数据。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "对于每个数据集（独热编码和未编码）：使用以下之一缩放所有非热编码值：StandardScaler、MinMaxScaler、MaxAbsScaler。\n",
    "比较在测试集上计算的误差确保计算偏斜（以确定是否应该进行转换）并仅在训练数据上拟合缩放器，然后将其同样应用于训练和测试数据\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:16: DeprecationWarning: `np.float` is a deprecated alias for the builtin `float`. To silence this warning, use `float` by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use `np.float64` here.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  app.launch_new_instance()\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "not_encoded - standardscaling        1.372182e+09\n",
      "not_encoded - minmaxscaling          1.372161e+09\n",
      "not_encoded - maxabsscaling          1.372683e+09\n",
      "one_hot_encoded - standardscaling    8.065328e+09\n",
      "one_hot_encoded - minmaxscaling      8.065328e+09\n",
      "one_hot_encoded - maxabsscaling      8.065328e+09\n",
      "--------------------------------------------------------------------------------\n",
      "not_encoded - standardscaling 1372182358.9345138\n",
      "not_encoded - minmaxscaling 1372160679.4641728\n",
      "not_encoded - maxabsscaling 1372682568.6731877\n",
      "one_hot_encoded - standardscaling 8065327607.2472315\n",
      "one_hot_encoded - minmaxscaling 8065327607.227574\n",
      "one_hot_encoded - maxabsscaling 8065327607.193239\n"
     ]
    }
   ],
   "source": [
    "from sklearn.preprocessing import StandardScaler, MinMaxScaler, MaxAbsScaler\n",
    "\n",
    "\n",
    "scalers = {'standard': StandardScaler(),\n",
    "           'minmax': MinMaxScaler(),\n",
    "           'maxabs': MaxAbsScaler()}\n",
    "\n",
    "training_test_sets = {\n",
    "    'not_encoded': (X_train, y_train, X_test, y_test),\n",
    "    'one_hot_encoded': (X_train_ohc, y_train_ohc, X_test_ohc, y_test_ohc)}\n",
    "\n",
    "\n",
    "# Get the list of float columns, and the float data\n",
    "# so that we don't scale something we already scaled. \n",
    "# We're supposed to scale the original data each time\n",
    "mask = X_train.dtypes == np.float\n",
    "float_columns = X_train.columns[mask]\n",
    "\n",
    "# 初始化模型\n",
    "LR = LinearRegression()\n",
    "\n",
    "# 迭代所有可能的联接获取errors\n",
    "errors = {}\n",
    "for encoding_label, (_X_train, _y_train, _X_test, _y_test) in training_test_sets.items():\n",
    "    for scaler_label, scaler in scalers.items():\n",
    "        trainingset = _X_train.copy()  # copy 因为不需要scale超过一次.\n",
    "        testset = _X_test.copy()\n",
    "        trainingset[float_columns] = scaler.fit_transform(trainingset[float_columns])\n",
    "        testset[float_columns] = scaler.transform(testset[float_columns])\n",
    "        LR.fit(trainingset, _y_train)\n",
    "        predictions = LR.predict(testset)\n",
    "        key = encoding_label + ' - ' + scaler_label + 'scaling'\n",
    "        errors[key] = mean_squared_error(_y_test, predictions)\n",
    "\n",
    "errors = pd.Series(errors)\n",
    "print(errors.to_string())\n",
    "print('-' * 80)\n",
    "for key, error_val in errors.items():\n",
    "    print(key, error_val)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "绘制其中一个模型的预测值与实际值"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "\n",
    "\n",
    "sns.set_context('talk')\n",
    "sns.set_style('ticks')\n",
    "sns.set_palette('dark')\n",
    "\n",
    "ax = plt.axes()\n",
    "# 使用 y_test, y_test_pred\n",
    "ax.scatter(y_test, y_test_pred, alpha=.5)\n",
    "\n",
    "ax.set(xlabel='Ground truth', \n",
    "       ylabel='Predictions',\n",
    "       title='Ames, Iowa House Price Predictions vs Truth, using Linear Regression');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['Alley', 'BldgType', 'BsmtCond', 'BsmtExposure', 'BsmtFinType1', 'BsmtFinType2', 'BsmtQual', 'CentralAir', 'Condition1', 'Condition2', 'Electrical', 'ExterCond', 'ExterQual', 'Exterior1st', 'Exterior2nd', 'Fence', 'FireplaceQu', 'Foundation', 'Functional', 'GarageCond', 'GarageFinish', 'GarageQual', 'GarageType', 'Heating', 'HeatingQC', 'HouseStyle', 'KitchenQual', 'LandContour', 'LandSlope', 'LotConfig', 'LotShape', 'MSZoning', 'MasVnrType', 'MiscFeature', 'Neighborhood', 'PavedDrive', 'PoolQC', 'RoofMatl', 'RoofStyle', 'SaleCondition', 'SaleType', 'Street', 'Utilities']\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:5: DeprecationWarning: `np.object` is a deprecated alias for the builtin `object`. To silence this warning, use `object` by itself. Doing this will not modify any behavior and is safe. \n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  \"\"\"\n"
     ]
    }
   ],
   "source": [
    "filepath = os.sep.join(data_path + ['Ames_Housing_Sales.csv'])\n",
    "data = pd.read_csv(filepath, sep=',')#加载数据\n",
    "\n",
    "# 获取由string类别组成的一个Pd.Series \n",
    "one_hot_encode_cols = data.dtypes[data.dtypes == np.object]  #  \n",
    "one_hot_encode_cols = one_hot_encode_cols.index.tolist()  # 离散值列表\n",
    "print(one_hot_encode_cols)\n",
    "for col in one_hot_encode_cols:\n",
    "    data[col] = pd.Categorical(data[col])\n",
    "\n",
    "#  one hot 编码\n",
    "data = pd.get_dummies(data, columns=one_hot_encode_cols)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "train, test = train_test_split(data, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "很多列由倾斜的特征-log转换可以应用"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:2: DeprecationWarning: `np.float` is a deprecated alias for the builtin `float`. To silence this warning, use `float` by itself. Doing this will not modify any behavior and is safe. If you specifically wanted the numpy scalar type, use `np.float64` here.\n",
      "Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdocs/release/1.20.0-notes.html#deprecations\n",
      "  \n"
     ]
    }
   ],
   "source": [
    "#创建一个值为float类型的列的列表来检查倾斜\n",
    "mask = data.dtypes == np.float\n",
    "float_cols = data.columns[mask]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Skew</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>MiscVal</th>\n",
       "      <td>26.915364</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>PoolArea</th>\n",
       "      <td>15.777668</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LotArea</th>\n",
       "      <td>11.501694</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LowQualFinSF</th>\n",
       "      <td>11.210638</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3SsnPorch</th>\n",
       "      <td>10.150612</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ScreenPorch</th>\n",
       "      <td>4.599803</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BsmtFinSF2</th>\n",
       "      <td>4.466378</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>EnclosedPorch</th>\n",
       "      <td>3.218303</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>LotFrontage</th>\n",
       "      <td>3.138032</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>MasVnrArea</th>\n",
       "      <td>2.492814</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>OpenPorchSF</th>\n",
       "      <td>2.295489</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>SalePrice</th>\n",
       "      <td>2.106910</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BsmtFinSF1</th>\n",
       "      <td>2.010766</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>TotalBsmtSF</th>\n",
       "      <td>1.979164</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1stFlrSF</th>\n",
       "      <td>1.539692</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GrLivArea</th>\n",
       "      <td>1.455564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>WoodDeckSF</th>\n",
       "      <td>1.334388</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>BsmtUnfSF</th>\n",
       "      <td>0.900308</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>GarageArea</th>\n",
       "      <td>0.838422</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2ndFlrSF</th>\n",
       "      <td>0.773655</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                    Skew\n",
       "MiscVal        26.915364\n",
       "PoolArea       15.777668\n",
       "LotArea        11.501694\n",
       "LowQualFinSF   11.210638\n",
       "3SsnPorch      10.150612\n",
       "ScreenPorch     4.599803\n",
       "BsmtFinSF2      4.466378\n",
       "EnclosedPorch   3.218303\n",
       "LotFrontage     3.138032\n",
       "MasVnrArea      2.492814\n",
       "OpenPorchSF     2.295489\n",
       "SalePrice       2.106910\n",
       "BsmtFinSF1      2.010766\n",
       "TotalBsmtSF     1.979164\n",
       "1stFlrSF        1.539692\n",
       "GrLivArea       1.455564\n",
       "WoodDeckSF      1.334388\n",
       "BsmtUnfSF       0.900308\n",
       "GarageArea      0.838422\n",
       "2ndFlrSF        0.773655"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "skew_limit = 0.75\n",
    "skew_vals = train[float_cols].skew()\n",
    "\n",
    "skew_cols = (skew_vals\n",
    "             .sort_values(ascending=False)\n",
    "             .to_frame()\n",
    "             .rename(columns={0:'Skew'})\n",
    "             .query('abs(Skew) > {0}'.format(skew_limit)))\n",
    "\n",
    "skew_cols"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "转换偏斜大于 0.75 的所有列，除了“SalePrice”。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# field = \"BsmtFinSF1\"\n",
    "# fig, (ax_before, ax_after) = plt.subplots(1, 2, figsize=(10, 5))\n",
    "# train[field].hist(ax=ax_before)\n",
    "# train[field].apply(np.log1p).hist(ax=ax_after)\n",
    "# ax_before.set(title='before np.log1p', ylabel='frequency', xlabel='value')\n",
    "# ax_after.set(title='after np.log1p', ylabel='frequency', xlabel='value')\n",
    "# fig.suptitle('Field \"{}\"'.format(field));"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:4: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  after removing the cwd from sys.path.\n",
      "H:\\Anaconda\\lib\\site-packages\\ipykernel_launcher.py:5: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
      "  \"\"\"\n"
     ]
    }
   ],
   "source": [
    "for col in skew_cols.index.tolist():\n",
    "    if col == \"SalePrice\":\n",
    "        continue\n",
    "    train[col] = np.log1p(train[col])\n",
    "    test[col]  = test[col].apply(np.log1p) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "feature_cols = [x for x in train.columns if x != 'SalePrice']\n",
    "X_train = train[feature_cols]\n",
    "y_train = train['SalePrice']\n",
    "\n",
    "X_test  = test[feature_cols]\n",
    "y_test  = test['SalePrice']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.metrics import mean_squared_error\n",
    "#计算均方差\n",
    "\n",
    "def rmse(ytrue, ypredicted):\n",
    "    return np.sqrt(mean_squared_error(ytrue, ypredicted))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "- 拟合基本线性回归模型\n",
    "- 打印此模型的均方根误差，\n",
    "- 绘制基于模型的预测与实际销售价格。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "306369.68342319806\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "linearRegression = LinearRegression().fit(X_train, y_train)\n",
    "\n",
    "linearRegression_rmse = rmse(y_test, linearRegression.predict(X_test))\n",
    "\n",
    "print(linearRegression_rmse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\cbook\\__init__.py:2064: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x[:, None]\n",
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\axes\\_base.py:248: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x = x[:, np.newaxis]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "f = plt.figure(figsize=(6,6))\n",
    "ax = plt.axes()\n",
    "\n",
    "ax.plot(y_test, linearRegression.predict(X_test), \n",
    "         marker='o', ls='', ms=3.0)\n",
    "\n",
    "lim = (0, y_test.max())\n",
    "\n",
    "ax.set(xlabel='Actual Price', \n",
    "       ylabel='Predicted Price', \n",
    "       xlim=lim,\n",
    "       ylim=lim,\n",
    "       title='Linear Regression Results');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    " "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "岭回归（翻看几个单元格后可以看到介绍）使用 L2 归一化来降低系数的大小。这在存在高方差的情况下会很有帮助。 \n",
    "Scikit-learn 中的正则化函数每个都包含内置交叉验证的版本。\n",
    "- 将常规（非交叉验证）岭模型拟合到一系列 α  值，并使用创建的交叉验证误差函数绘制 RMSE\n",
    "- 使用 [0.005,0.05,0.1,0.3,1,3,5,10,15,30,80] 作为 alpha 的范围。\n",
    "- 然后使用之前房价预测中的 α 值范围重复 Ridge 模型的拟合。比较结果。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15.0 32169.17620567246\n"
     ]
    }
   ],
   "source": [
    "#现在是 RidgeCV 方法。不幸的是，无法获得未选择模型的 alpha 值。由此产生的误差值和 α 值与上面获得的非常相似。\n",
    "from sklearn.linear_model import RidgeCV\n",
    "\n",
    "alphas = [0.005, 0.05, 0.1, 0.3, 1, 3, 5, 10, 15, 30, 80]\n",
    "\n",
    "ridgeCV = RidgeCV(alphas=alphas, \n",
    "                  cv=4).fit(X_train, y_train)\n",
    "\n",
    "ridgeCV_rmse = rmse(y_test, ridgeCV.predict(X_test))\n",
    "\n",
    "print(ridgeCV.alpha_, ridgeCV_rmse)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "与 RidgeCV 函数非常相似，还有一个 LassoCV 函数，它使用 L1 正则化函数和交叉验证。 \n",
    "L1 正则化会选择性地收缩一些系数，有效地执行特征消除。\n",
    "\n",
    "LassoCV 函数不允许设置评分函数。\n",
    "但是，上面创建的自定义误差函数 (rmse) 可用于评估最终模型的误差。\n",
    "\n",
    "同样，还有一个带有交叉验证的弹性网络函数 ElasticNetCV，它是 L2 和 L1 正则化的组合。\n",
    "\n",
    "- 使用交叉验证拟合 Lasso 模型，并使用上面创建的函数确定 α 和 RMSE 的最佳值。请注意，α 的大小可能与 Ridge 模型不同。\n",
    "- 对弹性网络模型重复此操作。\n",
    "- 通过表格和/或绘图比较结果。\n",
    "\n",
    "使用以下alphas:\n",
    "[1e-5, 5e-5, 0.0001, 0.0005]\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 94004559874.13617, tolerance: 387053725.5878631\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 93994604048.47357, tolerance: 387053725.5878631\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 93994499320.0592, tolerance: 387053725.5878631\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 93994441035.77199, tolerance: 387053725.5878631\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 107413893243.23529, tolerance: 391106099.75311756\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 107413079510.09709, tolerance: 391106099.75311756\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 107413353406.96588, tolerance: 391106099.75311756\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 107412835623.54077, tolerance: 391106099.75311756\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 120196147330.59464, tolerance: 442399396.3673559\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 117863439216.2968, tolerance: 442399396.3673559\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 117192989265.96474, tolerance: 442399396.3673559\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 117002595022.67313, tolerance: 442399396.3673559\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0005 39257.39399144776\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:532: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 200763773571.4656, tolerance: 610299085.7005944\n",
      "  positive)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import LassoCV\n",
    "\n",
    "alphas2 = np.array([1e-5, 5e-5, 0.0001, 0.0005])\n",
    "\n",
    "lassoCV = LassoCV(alphas=alphas2,\n",
    "                  max_iter=5e4,\n",
    "                  cv=3).fit(X_train, y_train)\n",
    "\n",
    "lassoCV_rmse = rmse(y_test, lassoCV.predict(X_test))\n",
    "\n",
    "print(lassoCV.alpha_, lassoCV_rmse) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Of 294 coefficients, 273 are non-zero with Lasso.\n"
     ]
    }
   ],
   "source": [
    "#查看非零值的特征\n",
    "print('Of {} coefficients, {} are non-zero with Lasso.'.format(len(lassoCV.coef_), \n",
    "                                                               len(lassoCV.coef_.nonzero()[0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 100160185887.15576, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 167868205507.94666, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 164350387911.20044, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162528161753.74152, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 19052201943.234467, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 140574092723.1702, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 137196981945.00302, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 134671321573.96957, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 54792800850.48314, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 145428933958.22775, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 143991180742.7182, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 143254030504.89203, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 46990355518.82254, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 181066000305.63336, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175377570092.8455, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 176173764777.16367, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 18476898522.825043, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 157217089080.71536, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 154430356351.66064, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152520064594.53494, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 116709813460.00856, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 167136354896.69672, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161487419961.40985, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162349452337.1952, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 32318521431.877686, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 139801747918.53964, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 136436135346.95409, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 134534657374.80215, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 82443198403.99467, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 145121450661.21976, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 143338957043.8362, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 143128546310.38956, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 78586857273.42131, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 180138867612.98312, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 174811028643.09116, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 176011212527.4145, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 54524536280.92325, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 156483163626.58362, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 153636181440.27957, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152385946304.10568, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 124290629034.23605, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 166399258362.76926, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162765289108.79517, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162153486963.73328, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 74890058674.08377, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 138996664556.63474, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 135172496280.03648, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 134373451952.15916, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 91905816577.12189, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 144807276704.13522, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 141924355856.68964, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142998997661.69662, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 91538578646.34306, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 179433796090.72025, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 173467572076.01324, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175831418998.15073, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 52913365371.18355, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 155703520461.01965, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152514795659.65662, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152234752668.78363, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142698539887.2403, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 165540844426.89914, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162715485974.85898, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161931683462.36642, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 78054852803.79121, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 138062641567.7824, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 132405085749.2714, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 134163857310.39546, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 112417508265.22238, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 144399770331.45743, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 141653855227.91492, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142899840796.75723, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 132467114992.26123, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 178483975577.79572, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 174040428746.33517, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175618822007.44757, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 77522813691.08354, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 154763849015.25214, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 150884407396.70532, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152055511737.5947, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 149820320685.15277, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 164346946005.687, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162424085238.4445, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161662097467.19308, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 113116488080.25784, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 137128920206.64261, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133666241783.10724, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133886416183.20755, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 135825254399.15845, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 143806105048.5022, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142502173678.367, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142795026663.8466, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 168167857855.10095, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 176990491397.21277, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 173842635677.20764, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175343558280.07684, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 140977247990.15524, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 153847495004.96246, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 150108811267.23776, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151840366061.941, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 157879737829.38727, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 163110809704.79102, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161330788396.4533, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161839632165.49274, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 130890384904.71245, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 135982096528.08821, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133479917546.97684, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133582905688.91739, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 141093181536.4247, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142791790937.3318, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 141814532957.4223, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142570744726.8622, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 174115296397.2028, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 173858151466.09198, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 172822928875.22913, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175420864418.0661, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 145787119962.16855, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152810044891.49902, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 150553392349.89365, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151589448906.59082, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162052070137.0011, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 162087751119.36395, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 159685610113.89508, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161934979653.41367, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 140048770387.18182, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 135298873426.20479, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133504622247.19885, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133300794169.04588, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142877388314.85318, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 141871637105.17136, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 140156764705.2618, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142140759989.9373, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 177289904137.20648, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 171538807209.4344, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 171553240216.89023, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175553040282.46747, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152430470051.98468, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 152357111340.85474, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151018354103.65833, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151763621518.19516, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 159557492115.19812, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 160754707534.8323, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 158923581821.64603, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161753910248.51907, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 139073962969.87222, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 133936563123.46048, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 132855856567.43324, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 132815865382.46504, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 144235019596.1983, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142393509718.35168, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 140974018636.57513, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142171345876.52625, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 179645310238.2392, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 173767187720.19122, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 168615226766.41034, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175379694571.73413, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 155726111271.003, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 149332881118.18604, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 149415597674.3957, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151715018415.92212, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 157428513144.58136, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 157764478211.0054, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 156948396880.35312, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 161267915202.54736, tolerance: 487372561.3038374\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 136163664527.45549, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 132547073357.64374, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 131638301943.83206, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 130983082258.54703, tolerance: 465192498.4087122\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 143997912656.8577, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142598199037.70316, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142214994295.44046, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 142042143617.12027, tolerance: 464569448.3765248\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 178215484695.9957, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 175133822856.65543, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 174320032066.7935, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 174909698135.13367, tolerance: 534955063.02047503\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 154028173018.09805, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151216771442.01492, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 150623935865.56747, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n",
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:528: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 151224892732.97787, tolerance: 488440735.7074999\n",
      "  tol, rng, random, positive)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0005 0.1 35001.234296074574\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:532: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 32162755250.822815, tolerance: 610299085.7005944\n",
      "  positive)\n"
     ]
    }
   ],
   "source": [
    "#现在尝试弹性网络，使用lasso同样的alpha,并且L1_ratios再0.1-0.9\n",
    "from sklearn.linear_model import ElasticNetCV\n",
    "\n",
    "l1_ratios = np.linspace(0.1, 0.9, 9)\n",
    "\n",
    "elasticNetCV = ElasticNetCV(alphas=alphas2, \n",
    "                            l1_ratio=l1_ratios,\n",
    "                            max_iter=1e4).fit(X_train, y_train)\n",
    "elasticNetCV_rmse = rmse(y_test, elasticNetCV.predict(X_test))\n",
    "\n",
    "print(elasticNetCV.alpha_, elasticNetCV.l1_ratio_, elasticNetCV_rmse)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>RMSE</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Linear</th>\n",
       "      <td>306369.683423</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Ridge</th>\n",
       "      <td>32169.176206</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lasso</th>\n",
       "      <td>39257.393991</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ElasticNet</th>\n",
       "      <td>35001.234296</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     RMSE\n",
       "Linear      306369.683423\n",
       "Ridge        32169.176206\n",
       "Lasso        39257.393991\n",
       "ElasticNet   35001.234296"
      ]
     },
     "execution_count": 86,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rmse_vals = [linearRegression_rmse, ridgeCV_rmse, lassoCV_rmse, elasticNetCV_rmse]\n",
    "\n",
    "labels = ['Linear', 'Ridge', 'Lasso', 'ElasticNet']\n",
    "\n",
    "rmse_df = pd.Series(rmse_vals, index=labels).to_frame()\n",
    "rmse_df.rename(columns={0: 'RMSE'}, inplace=1)\n",
    "rmse_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\cbook\\__init__.py:2064: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x[:, None]\n",
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\axes\\_base.py:248: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x = x[:, np.newaxis]\n",
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\cbook\\__init__.py:2064: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x[:, None]\n",
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\axes\\_base.py:248: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x = x[:, np.newaxis]\n",
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\cbook\\__init__.py:2064: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x[:, None]\n",
      "H:\\Anaconda\\lib\\site-packages\\matplotlib\\axes\\_base.py:248: FutureWarning: Support for multi-dimensional indexing (e.g. `obj[:, None]`) is deprecated and will be removed in a future version.  Convert to a numpy array before indexing instead.\n",
      "  x = x[:, np.newaxis]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# #作图\n",
    "# f = plt.figure(figsize=(6,6))\n",
    "# ax = plt.axes()\n",
    "\n",
    "# labels = ['Ridge', 'Lasso', 'ElasticNet']\n",
    "\n",
    "# models = [ridgeCV, lassoCV, elasticNetCV]\n",
    "\n",
    "# for mod, lab in zip(models, labels):\n",
    "#     ax.plot(y_test, mod.predict(X_test), \n",
    "#              marker='o', ls='', ms=3.0, label=lab)\n",
    "\n",
    "\n",
    "# leg = plt.legend(frameon=True)\n",
    "# leg.get_frame().set_edgecolor('black')\n",
    "# leg.get_frame().set_linewidth(1.0)\n",
    "\n",
    "# ax.set(xlabel='Actual Price', \n",
    "#        ylabel='Predicted Price', \n",
    "#        title='Linear Regression Results');"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "探索随机梯度下降。\n",
    "回想一下，线性模型通常对缩放敏感。然而，SGD 对缩放非常敏感。\n",
    "此外，学习率过高会导致算法发散，而学习率过低可能需要很长时间才能收敛。\n",
    "\n",
    "- 拟合一个没有正则化惩罚的随机梯度下降模型（相关参数是惩罚）。\n",
    "- 现在使用由上述交叉验证确定的参数值，用三个惩罚项（L2、L1、弹性网络）中的每一个拟合随机梯度下降模型。\n",
    "- 在拟合模型之前不要缩放数据。\n",
    "- 将结果与不使用随机梯度下降获得的结果进行比较。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>RMSE</th>\n",
       "      <th>RMSE-SGD</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Linear</th>\n",
       "      <td>306369.683423</td>\n",
       "      <td>5.359612e+15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Ridge</th>\n",
       "      <td>32169.176206</td>\n",
       "      <td>2.497940e+15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lasso</th>\n",
       "      <td>39257.393991</td>\n",
       "      <td>9.837581e+15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ElasticNet</th>\n",
       "      <td>35001.234296</td>\n",
       "      <td>3.809127e+15</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     RMSE      RMSE-SGD\n",
       "Linear      306369.683423  5.359612e+15\n",
       "Ridge        32169.176206  2.497940e+15\n",
       "Lasso        39257.393991  9.837581e+15\n",
       "ElasticNet   35001.234296  3.809127e+15"
      ]
     },
     "execution_count": 88,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#引入SGD回归器，准备参数\n",
    "from sklearn.linear_model import SGDRegressor\n",
    "\n",
    "model_parameters_dict = {\n",
    "    'Linear': {'penalty': 'none'},\n",
    "    'Lasso': {'penalty': 'l2',\n",
    "           'alpha': lassoCV.alpha_},\n",
    "    'Ridge': {'penalty': 'l1',\n",
    "           'alpha': ridgeCV_rmse},\n",
    "    'ElasticNet': {'penalty': 'elasticnet', \n",
    "                   'alpha': elasticNetCV.alpha_,\n",
    "                   'l1_ratio': elasticNetCV.l1_ratio_}\n",
    "}\n",
    "\n",
    "new_rmses = {}\n",
    "for modellabel, parameters in model_parameters_dict.items():\n",
    "    #将 dict项作为参数传递\n",
    "    SGD = SGDRegressor(**parameters)\n",
    "    SGD.fit(X_train, y_train)\n",
    "    new_rmses[modellabel] = rmse(y_test, SGD.predict(X_test))\n",
    "\n",
    "rmse_df['RMSE-SGD'] = pd.Series(new_rmses)\n",
    "rmse_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "注意误差值有多高！算法没有收敛。这可能是由于缩放和/或学习率太高。调整学习率，看看会发生什么\n",
    "\n",
    "- 创建SGDClassifier的实例时传入eta0=1e-7。\n",
    "- 重新计算所有惩罚的错误并进行比较。\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>RMSE</th>\n",
       "      <th>RMSE-SGD</th>\n",
       "      <th>RMSE-SGD-learningrate</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>Linear</th>\n",
       "      <td>306369.683423</td>\n",
       "      <td>5.359612e+15</td>\n",
       "      <td>74208.838374</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Ridge</th>\n",
       "      <td>32169.176206</td>\n",
       "      <td>2.497940e+15</td>\n",
       "      <td>74256.697993</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Lasso</th>\n",
       "      <td>39257.393991</td>\n",
       "      <td>9.837581e+15</td>\n",
       "      <td>74634.901623</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>ElasticNet</th>\n",
       "      <td>35001.234296</td>\n",
       "      <td>3.809127e+15</td>\n",
       "      <td>73245.998172</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                     RMSE      RMSE-SGD  RMSE-SGD-learningrate\n",
       "Linear      306369.683423  5.359612e+15           74208.838374\n",
       "Ridge        32169.176206  2.497940e+15           74256.697993\n",
       "Lasso        39257.393991  9.837581e+15           74634.901623\n",
       "ElasticNet   35001.234296  3.809127e+15           73245.998172"
      ]
     },
     "execution_count": 89,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.linear_model import SGDRegressor\n",
    "\n",
    "model_parameters_dict = {\n",
    "    'Linear': {'penalty': 'none'},\n",
    "    'Lasso': {'penalty': 'l2',\n",
    "           'alpha': lassoCV.alpha_},\n",
    "    'Ridge': {'penalty': 'l1',\n",
    "           'alpha': ridgeCV_rmse},\n",
    "    'ElasticNet': {'penalty': 'elasticnet', \n",
    "                   'alpha': elasticNetCV.alpha_,\n",
    "                   'l1_ratio': elasticNetCV.l1_ratio_}\n",
    "}\n",
    "\n",
    "new_rmses = {}\n",
    "for modellabel, parameters in model_parameters_dict.items():\n",
    "    # following notation passes the dict items as arguments\n",
    "    SGD = SGDRegressor(eta0=1e-7, **parameters)\n",
    "    SGD.fit(X_train, y_train)\n",
    "    new_rmses[modellabel] = rmse(y_test, SGD.predict(X_test))\n",
    "\n",
    "rmse_df['RMSE-SGD-learningrate'] = pd.Series(new_rmses)\n",
    "rmse_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "\n",
    "linear regression\n",
    "- 线性回归拟合别的数据。\n",
    "\n",
    "ridge regression\n",
    "\n",
    "lasso regression\n",
    "\n",
    "- 使用岭回归 (αalpha=0.001) 和套索回归 (α alpha=0.0001) 对使用具有多项式特征的别的数据进行回归。\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from __future__ import print_function\n",
    "import os\n",
    "data_path = ['dataset']\n",
    "\n",
    "filepath = os.sep.join(data_path + ['X_Y_Sinusoid_Data.csv'])\n",
    "data = pd.read_csv(filepath)#加载数据"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import PolynomialFeatures\n",
    "from sklearn.linear_model import LinearRegression\n",
    "\n",
    "# 设置多项式特征数量\n",
    "degree = 20\n",
    "pf = PolynomialFeatures(degree)\n",
    "lr = LinearRegression()#线性模型\n",
    "\n",
    "# 提取数据 \n",
    "X_data = data[['x']]\n",
    "#PolynomialFeatures 需要dataframe（具有一列，而不是series）或维度为 (X, 1) 的二维数组，其中 X 是长度。\n",
    "Y_data = data['y']\n",
    "\n",
    "# 多项式模型拟合数据\n",
    "X_poly = pf.fit_transform(X_data)\n",
    "lr = lr.fit(X_poly, Y_data)\n",
    "Y_pred = lr.predict(X_poly)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "H:\\Anaconda\\lib\\site-packages\\sklearn\\linear_model\\_coordinate_descent.py:532: ConvergenceWarning: Objective did not converge. You might want to increase the number of iterations. Duality gap: 0.4469570520796867, tolerance: 0.0012227586753890333\n",
      "  positive)\n"
     ]
    }
   ],
   "source": [
    "from sklearn.linear_model import Ridge, Lasso\n",
    "\n",
    "#   ridge 回归模型\n",
    "rr = Ridge(alpha=0.001)\n",
    "rr = rr.fit(X_poly, Y_data)\n",
    "Y_pred_rr = rr.predict(X_poly)\n",
    "\n",
    "#  lasso 回归模型\n",
    "lassor = Lasso(alpha=0.0001)\n",
    "lassor = lassor.fit(X_poly, Y_data)\n",
    "Y_pred_lr = lassor.predict(X_poly)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>linear regression</th>\n",
       "      <th>ridge regression</th>\n",
       "      <th>lasso regression</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>2.100000e+01</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>21.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>5.787139e+13</td>\n",
       "      <td>2.169397</td>\n",
       "      <td>2.167284</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>5.994042e+13</td>\n",
       "      <td>2.900278</td>\n",
       "      <td>4.706731</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>1.619308e+07</td>\n",
       "      <td>0.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>3.414772e+12</td>\n",
       "      <td>0.467578</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>3.634503e+13</td>\n",
       "      <td>1.017272</td>\n",
       "      <td>0.252181</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>1.064203e+14</td>\n",
       "      <td>2.883507</td>\n",
       "      <td>1.641353</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>1.643575e+14</td>\n",
       "      <td>12.429635</td>\n",
       "      <td>20.176708</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       linear regression  ridge regression  lasso regression\n",
       "count       2.100000e+01         21.000000         21.000000\n",
       "mean        5.787139e+13          2.169397          2.167284\n",
       "std         5.994042e+13          2.900278          4.706731\n",
       "min         1.619308e+07          0.000000          0.000000\n",
       "25%         3.414772e+12          0.467578          0.000000\n",
       "50%         3.634503e+13          1.017272          0.252181\n",
       "75%         1.064203e+14          2.883507          1.641353\n",
       "max         1.643575e+14         12.429635         20.176708"
      ]
     },
     "execution_count": 47,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 查看每个模型的绝对值系数\n",
    "\n",
    "coefficients = pd.DataFrame()\n",
    "coefficients['linear regression'] = lr.coef_.ravel()\n",
    "coefficients['ridge regression'] = rr.coef_.ravel()\n",
    "coefficients['lasso regression'] = lassor.coef_.ravel()\n",
    "coefficients = coefficients.applymap(abs)\n",
    "\n",
    "coefficients.describe()  # 正则化和非正则化数据差异很大"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}