[Project_001]_Machine_Learning_Concepts.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "[Project 001] Machine Learning Concepts.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "mount_file_id": "1xKYKB4C_tFvO3Nad2oGdcEH2rWvYdOPq",
      "authorship_tag": "ABX9TyPLSblpujHFWuu6lElDnbyL",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/id-shiv/knowledge_base/blob/master/%5BProject_001%5D_Machine_Learning_Concepts.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FwYi0CABMzDv",
        "colab_type": "text"
      },
      "source": [
        "# Concept"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nc5ClKiJEOCa",
        "colab_type": "text"
      },
      "source": [
        "**Traditional Method** \n",
        "\n",
        "```\n",
        "Input Data + Hardcoded Rules -> Output\n",
        "```\n",
        "\n",
        "**Machine Learning**\n",
        "\n",
        "```\n",
        "Training   : Input Data     + Output       -> General Rule \n",
        "  \n",
        "Prediction : New Input Data + General Rule -> Output (Predicted with Probability score)\n",
        "```"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8AxtsEh6Dy86",
        "colab_type": "text"
      },
      "source": [
        "# Example"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jHvJ1BcZq48h",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "* Input Data (Features) : Characteristics of the flower\n",
        "> * Sepal Length\n",
        "> * Sepal Width\n",
        "> * Petal Length\n",
        "> * Petal Width\n",
        "* Output (Label) : Class of Iris flower\n",
        "> * Versicolor\n",
        "> * Setosa\n",
        "> * Virginica"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-W80QOlmtISr",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=1XBfOJjt0tj7nQ0w3TdWhCwJRob2SMizy)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "T-ZuWRmdrM8r",
        "colab_type": "code",
        "outputId": "cbb0a839-cced-4de6-f6cc-24655952e445",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        }
      },
      "source": [
        "# Import Data\n",
        "url = \"https://raw.githubusercontent.com/jbrownlee/Datasets/master/iris.csv\"\n",
        "names = ['sepal-length', 'sepal-width', 'petal-length', 'petal-width', 'class']\n",
        "data = pd.read_csv(url, names=names)\n",
        "\n",
        "# Expected Output\n",
        "label = 'class'\n",
        "\n",
        "# View Data\n",
        "print(data.head())"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "   sepal-length  sepal-width  petal-length  petal-width        class\n",
            "0           5.1          3.5           1.4          0.2  Iris-setosa\n",
            "1           4.9          3.0           1.4          0.2  Iris-setosa\n",
            "2           4.7          3.2           1.3          0.2  Iris-setosa\n",
            "3           4.6          3.1           1.5          0.2  Iris-setosa\n",
            "4           5.0          3.6           1.4          0.2  Iris-setosa\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MLOx7QozEwrR",
        "colab_type": "code",
        "outputId": "c248c594-6e2e-45f8-800a-f8e4380f628c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        }
      },
      "source": [
        "# Characteristics\n",
        "print(f'Features : {features}')\n",
        "print(f'Label : {label}')\n",
        "print(f'Number of samples : {data.shape[0]}')\n",
        "print(f'Classes : {np.unique(data[label])}')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Features : ['sepal length (cm)', 'sepal width (cm)', 'petal length (cm)', 'petal width (cm)']\n",
            "Label : class\n",
            "Number of samples : 150\n",
            "Classes : ['Iris-setosa' 'Iris-versicolor' 'Iris-virginica']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xXk5Nlp8Ew1i",
        "colab_type": "code",
        "outputId": "f2ea5993-771a-4fb1-a5e7-17e87294ed6c",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        }
      },
      "source": [
        "# Samples\n",
        "print(np.array(data.loc[data['class'] == 'Iris-setosa'].iloc[0]))\n",
        "print(np.array(data.loc[data['class'] == 'Iris-versicolor'].iloc[0]))\n",
        "print(np.array(data.loc[data['class'] == 'Iris-virginica'].iloc[0]))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "[5.1 3.5 1.4 0.2 'Iris-setosa']\n",
            "[7.0 3.2 4.7 1.4 'Iris-versicolor']\n",
            "[6.3 3.3 6.0 2.5 'Iris-virginica']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wF75MSOqEN97",
        "colab_type": "text"
      },
      "source": [
        "# Terminologies"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yHInvr2-ERE_",
        "colab_type": "text"
      },
      "source": [
        "* Data Features : Attributes (Columns) of data that would used to determine the output.\n",
        "* Data Samples (X) : Each row in the data, set of features associated to a specific category.\n",
        "* Data Labels (y) : Output data to be predicted.\n",
        "- Class : Categories within Data Label\n",
        "* Data Characteristic : Univariate or Multivariate.\n",
        "* Attribute Characteristics : Data type (Integer, String, Date, etc.).\n",
        "* NaN : Missing values.\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qo1VqAWLkU8L",
        "colab_type": "text"
      },
      "source": [
        "# Workflow"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "FKzh9fctl3Lr",
        "colab_type": "text"
      },
      "source": [
        "* Data\n",
        "> * Import\n",
        "> * (If applicable) Explore Data\n",
        "> * Determine Problem\n",
        ">> * Data analysis?\n",
        ">> * Classification?\n",
        ">> * Regression?\n",
        ">> * Clustering?\n",
        "> * Pre-Process Data\n",
        ">> * Detect Features(X) and Labels (y)\n",
        ">> * (If applicable) Handle Missing values (NaN)\n",
        ">> * (If applicable) Remove \n",
        ">> * (If applicable) Augment Data\n",
        ">>> * (If applicable) Add new \\ Modify columns\n",
        ">> * (If applicable) Handle Non-Numeric Data\n",
        "* Model\n",
        "> * Determine  \n",
        "Based on Problem\n",
        "> * Create\n",
        "> * Split Data into Testing and Training Data\n",
        "> * Train  \n",
        "With Train Data\n",
        "> * Evaluate - Accuracy.  \n",
        "Evaluate the Model trained with Training Data against Testing Data.  \n",
        "Compare Actual Testing Data Output to Predicted Output\n",
        "> * (If applicable) Tune Model\n",
        "> * (If applicable) Save\n",
        "* Predict\n",
        "> * (If applicable) Load Model\n",
        "> * Predict"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ntKjFBb_LhbY",
        "colab_type": "text"
      },
      "source": [
        "# Libraries"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "6O_IOB9TLisH",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "import pandas as pd\n",
        "import numpy as np\n",
        "\n",
        "import sklearn\n",
        "from sklearn import linear_model, preprocessing, svm\n",
        "from sklearn import datasets, metrics\n",
        "from sklearn.neighbors import KNeighborsClassifier\n",
        "from sklearn.utils import shuffle\n",
        "from sklearn.cluster import KMeans\n",
        "from sklearn.datasets import load_iris\n",
        "from sklearn.preprocessing import scale\n",
        "\n",
        "import matplotlib.pyplot as pyplot\n",
        "from matplotlib import style\n",
        "\n",
        "import pickle"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yf9EYLINOSpR",
        "colab_type": "text"
      },
      "source": [
        "# Problem 1 - Regression with Linear Regression"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zfjFQBUKOVqw",
        "colab_type": "text"
      },
      "source": [
        "Predict G3 based on given data for all the students"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qz5H2bPsC6ZN",
        "colab_type": "text"
      },
      "source": [
        "### Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "415_RY5rCzCh",
        "colab_type": "code",
        "outputId": "f16d9a6a-28a8-40d2-f231-b34fcd65bbfd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        }
      },
      "source": [
        "data_file = '/content/drive/My Drive/050 Colaboratory Notebooks/Datasets/student-mat.csv'\n",
        "data = pd.read_csv(data_file, delimiter=';')\n",
        "data.columns"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Index(['school', 'sex', 'age', 'address', 'famsize', 'Pstatus', 'Medu', 'Fedu',\n",
              "       'Mjob', 'Fjob', 'reason', 'guardian', 'traveltime', 'studytime',\n",
              "       'failures', 'schoolsup', 'famsup', 'paid', 'activities', 'nursery',\n",
              "       'higher', 'internet', 'romantic', 'famrel', 'freetime', 'goout', 'Dalc',\n",
              "       'Walc', 'health', 'absences', 'G1', 'G2', 'G3'],\n",
              "      dtype='object')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 127
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kTC19eHdRIj9",
        "colab_type": "code",
        "outputId": "a4a280c1-4ea3-4ace-ee58-03e8ce4dda07",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 224
        }
      },
      "source": [
        "data = data[['G1', 'G2', 'G3', 'studytime', 'failures', 'absences']]\n",
        "data.head()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>G1</th>\n",
              "      <th>G2</th>\n",
              "      <th>G3</th>\n",
              "      <th>studytime</th>\n",
              "      <th>failures</th>\n",
              "      <th>absences</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>5</td>\n",
              "      <td>6</td>\n",
              "      <td>6</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>6</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>5</td>\n",
              "      <td>5</td>\n",
              "      <td>6</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>7</td>\n",
              "      <td>8</td>\n",
              "      <td>10</td>\n",
              "      <td>2</td>\n",
              "      <td>3</td>\n",
              "      <td>10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>15</td>\n",
              "      <td>14</td>\n",
              "      <td>15</td>\n",
              "      <td>3</td>\n",
              "      <td>0</td>\n",
              "      <td>2</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>6</td>\n",
              "      <td>10</td>\n",
              "      <td>10</td>\n",
              "      <td>2</td>\n",
              "      <td>0</td>\n",
              "      <td>4</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   G1  G2  G3  studytime  failures  absences\n",
              "0   5   6   6          2         0         6\n",
              "1   5   5   6          2         0         4\n",
              "2   7   8  10          2         3        10\n",
              "3  15  14  15          3         0         2\n",
              "4   6  10  10          2         0         4"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 128
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fsOJNtVgRcur",
        "colab_type": "text"
      },
      "source": [
        "## Pre-Process Data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7S41Frds0o0g",
        "colab_type": "text"
      },
      "source": [
        "### Create Features Vs Labels"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ymCNhJQvRpv4",
        "colab_type": "text"
      },
      "source": [
        "Set the label (Category \\ Classifier) feature of the data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "eceiVZpROJZS",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "label = 'G3'"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "w0UlGTIcRxlB",
        "colab_type": "text"
      },
      "source": [
        "Split Data into features (X) and labels (y)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yn_VFPWgOnM7",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "X = data.drop([label], axis=1)\n",
        "y = data[label]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "06-6kwI1v5rq",
        "colab_type": "text"
      },
      "source": [
        "## View Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "dY17t-x7IJuo",
        "colab_type": "code",
        "outputId": "f36838f1-dce0-4726-942d-a7c071464de1",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        }
      },
      "source": [
        "print(f'Number of Features : {len(X.columns)}')\n",
        "print(f'Features : {list(X.columns)}')\n",
        "print(f'Number of samples : {len(X)}')\n",
        "print(f'Number of classes : {len(np.unique(y))}')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of Features : 5\n",
            "Features : ['G1', 'G2', 'studytime', 'failures', 'absences']\n",
            "Number of samples : 395\n",
            "Number of classes : 18\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "n2jHJh0FPGKY",
        "colab_type": "text"
      },
      "source": [
        "## Split Train - Test"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZjOIIbnaq8ry",
        "colab_type": "text"
      },
      "source": [
        "Split Data into test and training samples,  \n",
        "This is to train the model with data different than that of what we would test the model with.\n",
        "\n",
        "Generally test to training sample ratio would be 10 to 20%"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "966VhDnDPZ1x",
        "colab_type": "text"
      },
      "source": [
        "Split the data as 4 arrays\n",
        "* X_train will be used to train the model to predict y_train\n",
        "* X_test will be used to test the prediction accuracy of the model against the actual y_test"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "a7F5l5uhO_Jy",
        "colab_type": "code",
        "outputId": "cb0c5331-0f62-41f1-98fe-9f5f95c2b0bd",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        }
      },
      "source": [
        "x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(X, y, test_size=0.1)\n",
        "print(len(x_train))\n",
        "print(len(y_train))\n",
        "print(len(x_test))\n",
        "print(len(y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "355\n",
            "355\n",
            "40\n",
            "40\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WKcXvlMcRUnJ",
        "colab_type": "text"
      },
      "source": [
        "## Linear Regression\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "2dVevsRrTKOy",
        "colab_type": "text"
      },
      "source": [
        "* Features plotted with N-dimentions. \n",
        "```\n",
        "N = Number of features\n",
        "```\n",
        "---\n",
        "Best Fit Line\n",
        "- Line that pierces through data points with minimum error\n",
        "- Defined by equation (for 2 dimentional data) : y = mx + c\n",
        "- m : slope of the line. Derived by taking 2 points in the slope and calculating it as: \n",
        "```\n",
        "m = ( y2 - y1 ) / ( x2 - x1 )\n",
        "```\n",
        "- b : is called the y intercept. i.e point in y-axis where the best fit line intercepts.\n",
        "---\n",
        "\n",
        "- Linear Regression may be used when there is a strong corelation between the data points. i.e. farthest point has minimum error.\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "e-GQnR3kUrcs",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=1K4stjFNSR1p9ucjBOojiALQYK0o90d03)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MP82JB3Gv8rg",
        "colab_type": "text"
      },
      "source": [
        "### Plot"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "URAoiM-vxXL0",
        "colab_type": "text"
      },
      "source": [
        "Plot X vs Label to visualize the data points and correlation\n",
        "\n",
        "Modify the x_label below to view correlation between different features and label"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iDI1eJ_Ev-d5",
        "colab_type": "code",
        "outputId": "172f091c-50f5-47cd-987e-681163ad8444",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        }
      },
      "source": [
        "x_label = 'G1'\n",
        "y_label = 'G3'\n",
        "\n",
        "x_points = data[x_label]\n",
        "y_points = data[y_label]\n",
        "\n",
        "style.use('ggplot')\n",
        "pyplot.scatter(x_points, y_points)\n",
        "pyplot.xlabel(x_label)\n",
        "pyplot.ylabel(y_label)\n",
        "\n",
        "pyplot.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEJCAYAAABlmAtYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "M81p8OkG1iZJ",
        "colab_type": "text"
      },
      "source": [
        "### Create Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wFoyMdhPawsN",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = linear_model.LinearRegression()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wgPMziC3c_si",
        "colab_type": "text"
      },
      "source": [
        "## Train"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MrPFDZ-cc_5V",
        "colab_type": "code",
        "outputId": "631f94c6-7d5b-4564-e4d1-89ed214af4c2",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "model.fit(x_train, y_train)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "LinearRegression(copy_X=True, fit_intercept=True, n_jobs=None, normalize=False)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 135
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "I506xP92dIld",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8Q3JC3rpdIuh",
        "colab_type": "code",
        "outputId": "87235c55-5362-43c5-918a-db9f6d59fc26",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "accuracy = model.score(x_test, y_test)\n",
        "print(accuracy)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.8298614153249448\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "i_suiWSGekfu",
        "colab_type": "text"
      },
      "source": [
        "## View"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QfrOBkbeelyk",
        "colab_type": "code",
        "outputId": "0d7b0f91-0bbc-4f5f-d988-2b5ba35662cc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        }
      },
      "source": [
        "print(f'Co-efficient (m) : {model.coef_}')\n",
        "print(f'Number of dimensions (N) : {len(model.coef_)}')\n",
        "print(f'Y-intercept (b) : {model.intercept_}')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Co-efficient (m) : [ 0.13346398  0.98977449 -0.20936944 -0.37186664  0.03273496]\n",
            "Number of dimensions (N) : 5\n",
            "Y-intercept (b) : -1.2969843058204322\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DKRZUgT3sKkm",
        "colab_type": "text"
      },
      "source": [
        "## Save"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "uOqmimKmsOVG",
        "colab_type": "text"
      },
      "source": [
        "Save the model as a pickle to re-training everytime the model is used for prediction"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZoNixKO9sU4r",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "with open('model.pickle', 'wb') as f:\n",
        "  pickle.dump(model, f)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "_yHPGEDDt2GH",
        "colab_type": "text"
      },
      "source": [
        "#### Save with best accuracy"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5MSDv1Njt7D9",
        "colab_type": "text"
      },
      "source": [
        "Loop N times to fit the model and save the model with highest accuracy between iterations\n",
        "\n",
        "Accuracy is different each time since split X-y is done randomly, i.e. training data may be different in each loop"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "m09hB8YRuH0z",
        "colab_type": "code",
        "outputId": "8cd936cc-55bf-424f-ae5e-d3f38c938a29",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 71
        }
      },
      "source": [
        "best_accuracy = 0\n",
        "least_accuracy = 100  # used only to see the difference in accuracy\n",
        "number_of_iterations = 50\n",
        "\n",
        "for _ in range(number_of_iterations):\n",
        "  x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(X, y, test_size=0.1)\n",
        "  model = linear_model.LinearRegression()\n",
        "  model.fit(x_train, y_train)\n",
        "  accuracy = model.score(x_test, y_test)\n",
        "\n",
        "  if accuracy > best_accuracy:\n",
        "    best_accuracy = accuracy\n",
        "    with open('model.pickle', 'wb') as f:\n",
        "      pickle.dump(model, f)\n",
        "  \n",
        "  if accuracy < least_accuracy:\n",
        "    least_accuracy = accuracy\n",
        "\n",
        "print(f'Best Accuracy : {best_accuracy}')\n",
        "print(f'Least Accuracy : {least_accuracy}')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Best Accuracy : 0.9312781709616988\n",
            "Least Accuracy : 0.5696106847107429\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U7GWaSx8sspG",
        "colab_type": "text"
      },
      "source": [
        "## Load"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "0IdH-eCEsthk",
        "colab_type": "text"
      },
      "source": [
        "Load the already trained model from the pickle file to be used for prediction"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AlvyOAi4szLh",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pickle_in = open('model.pickle', 'rb')\n",
        "model = pickle.load(pickle_in)\n",
        "pickle_in.close()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "YDbRAwyhf4Ob",
        "colab_type": "text"
      },
      "source": [
        "## Predict"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mOEs9Egnf6VD",
        "colab_type": "code",
        "outputId": "7f4d409d-d335-4517-9b8b-3dbd0ac0d1fc",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 887
        }
      },
      "source": [
        "predictions = model.predict(x_test)\n",
        "\n",
        "# Predict for 5 samples from test data\n",
        "for x in range(10):\n",
        "  print(f'Input Data : {x_test.to_numpy()[x]}')\n",
        "  print(f'Prediction : {int(round(predictions[x]))}')\n",
        "  print(f'Actual : {y_test.to_numpy()[x]}')\n",
        "  print('--------------\\n')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Input Data : [11  9  1  1 19]\n",
            "Prediction : 9\n",
            "Actual : 10\n",
            "--------------\n",
            "\n",
            "Input Data : [ 9 11  3  0  2]\n",
            "Prediction : 10\n",
            "Actual : 11\n",
            "--------------\n",
            "\n",
            "Input Data : [17 18  1  0  0]\n",
            "Prediction : 19\n",
            "Actual : 18\n",
            "--------------\n",
            "\n",
            "Input Data : [10 10  1  0  4]\n",
            "Prediction : 10\n",
            "Actual : 10\n",
            "--------------\n",
            "\n",
            "Input Data : [5 9 2 2 6]\n",
            "Prediction : 7\n",
            "Actual : 7\n",
            "--------------\n",
            "\n",
            "Input Data : [7 9 3 0 0]\n",
            "Prediction : 8\n",
            "Actual : 8\n",
            "--------------\n",
            "\n",
            "Input Data : [11 12  2  0 12]\n",
            "Prediction : 12\n",
            "Actual : 11\n",
            "--------------\n",
            "\n",
            "Input Data : [11  9  2  0 12]\n",
            "Prediction : 9\n",
            "Actual : 9\n",
            "--------------\n",
            "\n",
            "Input Data : [8 6 2 2 2]\n",
            "Prediction : 5\n",
            "Actual : 5\n",
            "--------------\n",
            "\n",
            "Input Data : [9 9 2 0 6]\n",
            "Prediction : 8\n",
            "Actual : 10\n",
            "--------------\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QgS5VvmYDm5h",
        "colab_type": "text"
      },
      "source": [
        "# Problem 2 - Classification with K-Neighbors"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZRS303LyGPkX",
        "colab_type": "text"
      },
      "source": [
        "## Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xkgwZ0kvFPGw",
        "colab_type": "code",
        "outputId": "69cc748e-4280-498b-8caf-074064c563b0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "data_file = 'https://archive.ics.uci.edu/ml/machine-learning-databases/car/car.data'\n",
        "data = pd.read_csv(data_file, names=['buying', 'maintenance', 'doors', 'persons', 'lug_boot', 'safety', 'class'])\n",
        "print(data.shape)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(1728, 7)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WagaK_JkGhSe",
        "colab_type": "text"
      },
      "source": [
        "## Pre-Process Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N0V7pEzPw9UR",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "label = 'class'\n",
        "features = ['buying', 'maintenance', 'doors', 'persons', 'lug_boot', 'safety']\n",
        "classes = np.unique(data[label])"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "84m-4HNJttA-",
        "colab_type": "text"
      },
      "source": [
        "### Create Features Vs Labels"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "N_ihB9JSxHql",
        "colab_type": "text"
      },
      "source": [
        "Encode text to numeric data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "1dwb4Rwmt1eL",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "encoder = preprocessing.LabelEncoder()\n",
        "\n",
        "X = pd.DataFrame()\n",
        "y = list(encoder.fit_transform(list(data[label])))\n",
        "\n",
        "for feature in features:\n",
        "  X[feature] = encoder.fit_transform(list(data[feature]))\n",
        "X = X.to_numpy()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BPWGJam7G8bT",
        "colab_type": "text"
      },
      "source": [
        "## View Data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Y8xu2wAAHFJf",
        "colab_type": "code",
        "outputId": "61dd3026-093d-428d-9874-d294ceb50e6a",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 139
        }
      },
      "source": [
        "print(f'Number of Features : {len(X[0])}')\n",
        "print(f'Features : {features}')\n",
        "print(f'Label : {label}')\n",
        "print(f'Number of samples : {len(X)}')\n",
        "print(f'Number of classes : {len(classes)}')\n",
        "print(f'Classes : {classes}')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Number of Features : 6\n",
            "Features : ['buying', 'maintenance', 'doors', 'persons', 'lug_boot', 'safety']\n",
            "Label : class\n",
            "Number of samples : 1728\n",
            "Number of classes : 4\n",
            "Classes : ['acc' 'good' 'unacc' 'vgood']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1ps1ALjAzwys",
        "colab_type": "text"
      },
      "source": [
        "## Split Train - Test"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "gTcd03byzzEU",
        "colab_type": "code",
        "outputId": "968d8430-b50e-4b08-f461-34a69abf2926",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 105
        }
      },
      "source": [
        "x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(X, y, test_size=0.1)\n",
        "print(len(x_train))\n",
        "print(len(y_train))\n",
        "print(len(x_test))\n",
        "print(len(y_test))"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "1555\n",
            "1555\n",
            "173\n",
            "173\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "f_lxDZI90SNn",
        "colab_type": "text"
      },
      "source": [
        "## K-Nearest Neighbor (KNN)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "K5mWg0BE0aJn",
        "colab_type": "text"
      },
      "source": [
        "Introduction:\n",
        "- Works on classification problem where\n",
        "- data points are grouped into clusters of different class from input data.\n",
        "- Attempts to classify given data point to one of the classes.\n",
        "---\n",
        "Concept :  \n",
        "\n",
        "- K is a hyper parameter, is the number of closest neighboring data points that the algorithm would look for from the input data point.\n",
        "- The identified K neighboring data points would then vote their class.\n",
        "- More the votes from a particular class, input data point would be classified as belonging to that class\n",
        "---\n",
        "Best Practices : \n",
        "- Pick an odd number for K to have a decider vote in case of a tie\n",
        "---\n",
        "Math :  \n",
        "- Find the magnitude (measured using Eucliedian Distance) of the line between input data point and trained data points\n",
        "- For 2-dimentional data point:\n",
        "```\n",
        "Equlidian Distance (d) = Square_Root(square(x2 - x1) + square(y2 - y1))\n",
        "```\n",
        "- Pick K points who's magnitude is the highest"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "stApm6lZTi51",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=1N7pDWWPMjDDYqraE8Svjog3Vf5JOVfBH)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AmP2zWIP164W",
        "colab_type": "text"
      },
      "source": [
        "### Plot"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "LOsphFlI1-s2",
        "colab_type": "text"
      },
      "source": [
        "Plot X vs Label to visualize the data points and correlation\n",
        "\n",
        "Modify the x_label below to view correlation between different features and label"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R3nGSrl512Fs",
        "colab_type": "code",
        "outputId": "0ce5d945-aa3c-4568-eab8-73e671f5e4d8",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 302
        }
      },
      "source": [
        "x_label = label\n",
        "y_label = 'safety'\n",
        "\n",
        "x_points = data[x_label]\n",
        "y_points = data[y_label]\n",
        "\n",
        "style.use('ggplot')\n",
        "pyplot.scatter(x_points, y_points)\n",
        "pyplot.xlabel(x_label)\n",
        "pyplot.ylabel(y_label)\n",
        "\n",
        "pyplot.show()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEJCAYAAABlmAtYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1CChn7sO66Aj",
        "colab_type": "text"
      },
      "source": [
        "### Create Model"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "mTY_6bM367rX",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "k = 7\n",
        "model = KNeighborsClassifier(n_neighbors=k)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ut75EmVf7G2I",
        "colab_type": "text"
      },
      "source": [
        "## Train"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GKWWL7vn7I9z",
        "colab_type": "code",
        "outputId": "6a9f9f1b-6838-4818-cd49-8a5e14a9ef12",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        }
      },
      "source": [
        "model.fit(x_train, y_train)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',\n",
              "                     metric_params=None, n_jobs=None, n_neighbors=7, p=2,\n",
              "                     weights='uniform')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 149
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GsL6l7nT7Ril",
        "colab_type": "text"
      },
      "source": [
        "## Evaluate"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "czGEysEZ7Svw",
        "colab_type": "code",
        "outputId": "e93e8ce8-efe1-4009-ab55-6ba06ab99fde",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "accuracy = model.score(x_test, y_test)\n",
        "print(accuracy)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0.953757225433526\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "d8-bK-3m7jd8",
        "colab_type": "text"
      },
      "source": [
        "## Predict"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Z_X76Y437kY8",
        "colab_type": "code",
        "outputId": "91b09169-a69a-4bd4-c887-8e2dde6533e4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 887
        }
      },
      "source": [
        "predictions = model.predict(x_test)\n",
        "\n",
        "# Predict for 5 samples from test data\n",
        "for x in range(10):\n",
        "  print(f'Input Data : {x_test[x]}')\n",
        "  print(f'Prediction : {int(round(predictions[x]))}')\n",
        "  # print(f'Neighbors : {model.kneighbors([x_test[x]], n_neighbors=k, return_distance=True)}')\n",
        "  print(f'Actual : {y_test[x]}')\n",
        "  print('--------------\\n')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Input Data : [1 3 3 1 1 1]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [1 3 3 0 1 1]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [2 0 0 0 1 1]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [3 1 0 0 0 2]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [2 1 2 2 1 0]\n",
            "Prediction : 3\n",
            "Actual : 3\n",
            "--------------\n",
            "\n",
            "Input Data : [2 2 3 2 2 2]\n",
            "Prediction : 0\n",
            "Actual : 0\n",
            "--------------\n",
            "\n",
            "Input Data : [0 0 1 1 1 1]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [3 3 1 1 2 2]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [0 1 1 0 0 2]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n",
            "Input Data : [0 3 1 1 2 1]\n",
            "Prediction : 2\n",
            "Actual : 2\n",
            "--------------\n",
            "\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "B4W6Cv8PN3pn",
        "colab_type": "text"
      },
      "source": [
        "# Problem 3 - Classification with SVM"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JmITJLI0OJoO",
        "colab_type": "text"
      },
      "source": [
        "## Support Vector Machines (SVM)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jCmzhWvxONfp",
        "colab_type": "text"
      },
      "source": [
        "- Split clusters of data points by hyper-plane (*center line in below image*)\n",
        "- Difference between the distance from hyper-plane to nearest data points (also called **Support Vectors**) in either clusters are very minimal or zero\n",
        "- Number of hyper-planes may be number of clusters minus 1\n",
        "---\n",
        "How to choose hyper-plane?\n",
        "- Margine has to be maximum\n",
        "- What is Margin? it is area between two lines parallel to hyper-plane (*dotted lines in below image*) drawn on either side of the hyper-plane and seperating out the data points cluster"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "g0AdW4PZQYjz",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=13txs1xAkJr7Dsp5XTyZsxmqwfNVAgikf)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "in4fdXGZTHv-",
        "colab_type": "text"
      },
      "source": [
        "**When hyper-plane cannot clearly distinguish the data points clusters?**"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "k4ZwN1lsVJYb",
        "colab_type": "text"
      },
      "source": [
        "**Solution** : Converting N dimentional data point to N + 1 dimentional data point using Kernel\n",
        "\n",
        "**Kernel** : is a function that takes a data point in N-dimensions to outputs N+1th co-ordinate of the data point"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "iti5CPjdWQf2",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=1zxssICVmwoJPh8UqrX1NxVQREpxLAo7V)"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "tlUjcwehXeGz",
        "colab_type": "code",
        "outputId": "6969f66a-41d4-4c7a-b4f4-3f3a6efcdf33",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 54
        }
      },
      "source": [
        "# Cancer data set from sklearn\n",
        "data = datasets.load_breast_cancer()\n",
        "\n",
        "X = data.data\n",
        "y = data.target\n",
        "\n",
        "x_train, x_test, y_train, y_test = sklearn.model_selection.train_test_split(X, y, test_size=0.1)\n",
        "\n",
        "class_names = ['malignant', 'benign']\n",
        "classes = np.unique(y)\n",
        "\n",
        "best_fit_kernel = ''\n",
        "best_accuracy = 0\n",
        "for kernel in ['rbf', 'linear', 'poly']:\n",
        "  model = svm.SVC(kernel=kernel)\n",
        "  model.fit(x_train, y_train)\n",
        "\n",
        "  y_prediction = model.predict(x_test)\n",
        "\n",
        "  accuracy = metrics.accuracy_score(y_test, y_prediction)\n",
        "\n",
        "  if accuracy > best_accuracy:\n",
        "    best_accuracy = accuracy\n",
        "    best_fit_kernel = kernel\n",
        "print(f'Best fit kernel is {best_fit_kernel} with accuracy {best_accuracy}')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Best fit kernel is linear with accuracy 0.9649122807017544\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zNXUqsbhaOgt",
        "colab_type": "text"
      },
      "source": [
        "# Problem 4 - Clustering with K-Means"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8kUzj9TWIJRW",
        "colab_type": "text"
      },
      "source": [
        "- K : # of clusters\n",
        "- Centroid : X (in the image below) is the center point to the data point"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "92fs9s3ManR6",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=1Cilzuyhl-AE_9Wlg6ERJwhP0_5B_4ALZ)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P6T5rTkMJRQe",
        "colab_type": "text"
      },
      "source": [
        "**Working**  \n",
        "Start by placing K centroids randomly in the plot.\n",
        "- Draw a line between the centroids\n",
        "- Draw another line perpendicular to the above line at the center of above line\n",
        "- Consider all data points inside of the line as associated to that centroid i.e. cluster\n",
        "- Move the centroids to the center of the cluster by generation data point of centroid as average of all data points belonging to that centroid\n",
        "- Repeat above above untill no changes in data points assignment to centroid. i.e. cluster"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qykDQ0HBIt-x",
        "colab_type": "text"
      },
      "source": [
        "## Data"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H2UXFJLuVDRh",
        "colab_type": "text"
      },
      "source": [
        "Load iris dataset from sklearn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JdlsA2U5UYvY",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "data = load_iris()\n",
        "\n",
        "# For computation reasons, scale features within the value -1 and 1\n",
        "X = scale(data.data)\n",
        "y = pd.DataFrame(data.target)\n",
        "\n",
        "features = data.feature_names\n",
        "\n",
        "k = 3"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dRLFsA5DxQuv",
        "colab_type": "text"
      },
      "source": [
        "## Model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z78HahQXxVVE",
        "colab_type": "text"
      },
      "source": [
        "### Score"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aMFtPWHjxUp0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "def bench_k_means(estimator, name, data):\n",
        "    estimator.fit(data)\n",
        "    print('%-9s\\t%i\\t%.3f\\t%.3f\\t%.3f\\t%.3f\\t%.3f\\t%.3f'\n",
        "          % (name, estimator.inertia_,\n",
        "             metrics.homogeneity_score(y, estimator.labels_),\n",
        "             metrics.completeness_score(y, estimator.labels_),\n",
        "             metrics.v_measure_score(y, estimator.labels_),\n",
        "             metrics.adjusted_rand_score(y, estimator.labels_),\n",
        "             metrics.adjusted_mutual_info_score(y,  estimator.labels_),\n",
        "             metrics.silhouette_score(data, estimator.labels_,\n",
        "                                      metric='euclidean')))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "1BtyRYImxc6K",
        "colab_type": "text"
      },
      "source": [
        "### Train"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3RU-mGiAxfOD",
        "colab_type": "code",
        "outputId": "4e9a1937-59a1-49c9-b21e-7a2bd3b0ee9e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        }
      },
      "source": [
        "model = KMeans(n_clusters=k, init=\"random\", n_init=10)\n",
        "# bench_k_means(model, \"1\", data)\n",
        "model.fit(X)"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "KMeans(algorithm='auto', copy_x=True, init='random', max_iter=300, n_clusters=3,\n",
              "       n_init=10, n_jobs=None, precompute_distances='auto', random_state=None,\n",
              "       tol=0.0001, verbose=0)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 155
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Zp0m9-N3fax0",
        "colab_type": "text"
      },
      "source": [
        "### View"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hu85Wwflfbvn",
        "colab_type": "code",
        "outputId": "4a96ac3c-9af3-46e7-b489-6dd967509d92",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 319
        }
      },
      "source": [
        "X_df = pd.DataFrame(data.data)\n",
        "X_df.columns = features\n",
        "y.columns = ['Type']\n",
        "\n",
        "color_theme = np.array(['darkgrey', 'lightsalmon', 'powderblue'])\n",
        "\n",
        "pyplot.subplot(1, 2, 1)\n",
        "pyplot.scatter(x = X_df[X_df.columns[0]], \n",
        "               y = X_df[X_df.columns[1]],\n",
        "               c=color_theme[data.target],\n",
        "               s=50)\n",
        "pyplot.title('Actual')\n",
        "\n",
        "pyplot.subplot(1, 2, 2)\n",
        "pyplot.scatter(x = X_df[X_df.columns[0]], \n",
        "               y = X_df[X_df.columns[1]],\n",
        "               c=color_theme[model.labels_],\n",
        "               s=50)\n",
        "pyplot.title('Predicted')"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Text(0.5, 1.0, 'Predicted')"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 156
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEJCAYAAACE39xMAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBaMAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<Figure size 432x288 with 2 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZULLWKIarlOJ",
        "colab_type": "text"
      },
      "source": [
        "# All Models"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qjPTLE-srnn8",
        "colab_type": "text"
      },
      "source": [
        "![](https://drive.google.com/uc?export=view&id=1PAO2MwbcIAkSjf7PqcZEOZ55_1Raw7ej)"
      ]
    }
  ]
}