04-kNN/05-Hyper-Parameters/05-Hyper-Parameters.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 05 超参数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from sklearn import datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "digits = datasets.load_digits()\n",
    "X = digits.data\n",
    "y = digits.target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=666)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.98888888888888893"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "\n",
    "knn_clf = KNeighborsClassifier(n_neighbors=3)\n",
    "knn_clf.fit(X_train, y_train)\n",
    "knn_clf.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 寻找最好的k"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "best_k = 4\n",
      "best_score = 0.991666666667\n"
     ]
    }
   ],
   "source": [
    "best_score = 0.0\n",
    "best_k = -1\n",
    "for k in range(1, 11):\n",
    "    knn_clf = KNeighborsClassifier(n_neighbors=k)\n",
    "    knn_clf.fit(X_train, y_train)\n",
    "    score = knn_clf.score(X_test, y_test)\n",
    "    if score > best_score:\n",
    "        best_k = k\n",
    "        best_score = score\n",
    "        \n",
    "print(\"best_k =\", best_k)\n",
    "print(\"best_score =\", best_score)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 考虑距离？不考虑距离？"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "best_method = uniform\n",
      "best_k = 4\n",
      "best_score = 0.991666666667\n"
     ]
    }
   ],
   "source": [
    "best_score = 0.0\n",
    "best_k = -1\n",
    "best_method = \"\"\n",
    "for method in [\"uniform\", \"distance\"]:\n",
    "    for k in range(1, 11):\n",
    "        knn_clf = KNeighborsClassifier(n_neighbors=k, weights=method)\n",
    "        knn_clf.fit(X_train, y_train)\n",
    "        score = knn_clf.score(X_test, y_test)\n",
    "        if score > best_score:\n",
    "            best_k = k\n",
    "            best_score = score\n",
    "            best_method = method\n",
    "        \n",
    "print(\"best_method =\", best_method)\n",
    "print(\"best_k =\", best_k)\n",
    "print(\"best_score =\", best_score)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.98333333333333328"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sk_knn_clf = KNeighborsClassifier(n_neighbors=4, weights=\"distance\", p=1)\n",
    "sk_knn_clf.fit(X_train, y_train)\n",
    "sk_knn_clf.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 搜索明可夫斯基距离相应的p"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "best_k = 3\n",
      "best_p = 2\n",
      "best_score = 0.988888888889\n"
     ]
    }
   ],
   "source": [
    "best_score = 0.0\n",
    "best_k = -1\n",
    "best_p = -1\n",
    "\n",
    "for k in range(1, 11):\n",
    "    for p in range(1, 6):\n",
    "        knn_clf = KNeighborsClassifier(n_neighbors=k, weights=\"distance\", p=p)\n",
    "        knn_clf.fit(X_train, y_train)\n",
    "        score = knn_clf.score(X_test, y_test)\n",
    "        if score > best_score:\n",
    "            best_k = k\n",
    "            best_p = p\n",
    "            best_score = score\n",
    "        \n",
    "print(\"best_k =\", best_k)\n",
    "print(\"best_p =\", best_p)\n",
    "print(\"best_score =\", best_score)"
   ]
  }
 ],
 "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}