ML_Lab9.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "ML_Lab9.ipynb",
      "provenance": [],
      "authorship_tag": "ABX9TyOvNIgxlgShB95kNjsCZOuI",
      "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/JARACH-209/AI-ML_Lab/blob/master/ML_Lab9.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DtXl-Lh-A33h"
      },
      "source": [
        "##Achal Dixit\n",
        "1801007"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "10xKimT3Ax2w"
      },
      "source": [
        "import numpy as np\n",
        "import pandas as pd\n",
        "import matplotlib.pyplot as plt\n",
        "from sklearn.datasets import load_iris\n",
        "import math\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix\n",
        "%matplotlib inline\n",
        "\n",
        "X,Y = load_iris(return_X_y=True)\n",
        "df = pd.DataFrame(X)"
      ],
      "execution_count": 75,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "k_s_oWq3BGuZ",
        "outputId": "7bcd0c41-b3ba-40cb-9946-1c643373c564",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 284
        }
      },
      "source": [
        "plt.style.use(\"ggplot\")\n",
        "plt.scatter(df[2][50:],df[3][50:],c=Y[50:])\n",
        "# Versicolor and Virginica with features Petal Length and width\n",
        "# Non-Linearly Separable "
      ],
      "execution_count": 72,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.collections.PathCollection at 0x7f8b1c17e390>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 72
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "cFUZ-LEMLwu6"
      },
      "source": [
        "df.drop(columns=[0,1],inplace = True)"
      ],
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0sroSpNCL47h"
      },
      "source": [
        "# Taking only two classes and two features\n",
        "X = df.iloc[50:].to_numpy()\n",
        "Y = Y[50:]"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nagXA_vvOA44"
      },
      "source": [
        "df_s = pd.DataFrame(X)\n",
        "df_s.loc[df_s[1] == 1.5]    #point 56, 69, 79, 83, and 84 are causing problem "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hQgxUcBJPiK_"
      },
      "source": [
        "df_s.drop(index = [56,69,79,83,84],inplace= True)\n",
        "liers = [56,69,79,83,84]\n",
        "Y = np.delete(Y,liers)"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "sIE4w7WSQ5Ar"
      },
      "source": [
        "X = df_s.to_numpy()"
      ],
      "execution_count": 6,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lq9FmIfqQexP",
        "outputId": "df9bde3d-e887-4d91-fa80-d5522ea6b02e",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 284
        }
      },
      "source": [
        "plt.style.use(\"ggplot\")\n",
        "plt.scatter(df_s[0],df_s[1],c=Y)\n",
        "# Dataset is linearly separable"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.collections.PathCollection at 0x7fec6ff0e978>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "aYEpC2p9B82w"
      },
      "source": [
        "from sklearn.model_selection import train_test_split\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "\n",
        "x_train, x_test,y_train,y_test = train_test_split(X,Y,test_size = 0.2)\n",
        "scaler = StandardScaler().fit(x_train)\n",
        "x_train = scaler.transform(x_train)\n",
        "scaler = StandardScaler().fit(x_test)\n",
        "x_test = scaler.transform(x_test)"
      ],
      "execution_count": 12,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ywjwlW9DH1S6"
      },
      "source": [
        "from sklearn.model_selection import GridSearchCV\n",
        "from sklearn.svm import SVC\n",
        "from sklearn.metrics import accuracy_score, confusion_matrix\n",
        "\n",
        "svc = SVC()\n",
        "parameter =  {'kernel':('linear', 'rbf')}\n",
        "clf = GridSearchCV(svc,parameter,cv = 5)  #5-fold CV\n",
        "clf.fit(x_train,y_train)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bftRXdCtDy99"
      },
      "source": [
        "y_pred = clf.predict(x_test)\n",
        "matrix = confusion_matrix(y_test, y_pred)\n",
        "class_acc = matrix.diagonal()/matrix.sum(axis=1)"
      ],
      "execution_count": 108,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "SG7qTuWIF7mM",
        "outputId": "49b09049-3e54-41ed-c148-b04da795ce88",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "print(\"Accuracy Overall \\t\\t%0.2f\"%(accuracy_score(y_test,y_pred)*100))\n",
        "print(\"Accuracy Class Versicolor \\t%0.2f\"%(class_acc[0]*100))\n",
        "print(\"Accuracy Class Virginica \\t%0.2f\"%(class_acc[1]*100))"
      ],
      "execution_count": 109,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy Overall \t\t94.74\n",
            "Accuracy Class Versicolor \t91.67\n",
            "Accuracy Class Virginica \t100.00\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7uAw5B4j-Ykx"
      },
      "source": [
        "RBFANN"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DHSrmMDt-a6d"
      },
      "source": [
        "import math\n",
        "class K_Means():\n",
        "    def __init__(self,k=1,iterations = 100,elbow=False,k_max=10):\n",
        "        self.k = k\n",
        "        self.iterations = iterations\n",
        "        self.elbow_flag = elbow\n",
        "        self.k_max = k_max\n",
        "        self.best_k = None\n",
        "        self.elbow_reductions = None\n",
        "        self.k_centroids = None\n",
        "        self.y = None\n",
        "        self.variances = None\n",
        "\n",
        "    def final_centroids(self):\n",
        "        return np.array(self.k_centroids)\n",
        "\n",
        "    def get_variance(self,X,y,k_centroids):\n",
        "        # sum of squares of distances between points and their centroids\n",
        "        temp = 0\n",
        "        for j in range(len(k_centroids)):\n",
        "            for point,cluster in zip(X,y):\n",
        "                if cluster == j:\n",
        "                    temp += self.euclidean_distance(point,k_centroids[j])**2 \n",
        "        return math.sqrt(temp)\n",
        "\n",
        "    def euclidean_distance(self, p1, p2):\n",
        "        # math.dist can gives distance between points of any dimensionality \n",
        "        # Unfortunately Colab works on python3.6 and dist is in 3.8\n",
        "        #return math.dist(p1,p2)     \n",
        "        return np.linalg.norm(p1-p2)\n",
        "\n",
        "    def get_centroids(self,X,y,k):\n",
        "        k_centroids = []\n",
        "        cluster_indexes = []\n",
        "        for j in range(k):\n",
        "            cluster_indexes.append([i for i in range(len(y)) if y[i] == j])\n",
        "        # Finding means of all the clusters\n",
        "        # Summing all the coordinates in all the dimensions having same cluster \n",
        "        # label and taking a mean\n",
        "        for cluster in cluster_indexes:\n",
        "            cluster_mean = 0\n",
        "            for i in cluster:\n",
        "                 cluster_mean += X[i]\n",
        "            num = len(cluster) if len(cluster) >0 else 1\n",
        "            k_centroids.append(np.array(cluster_mean/num))\n",
        "            \n",
        "            # k = number of clusters and their labels as well.\n",
        "            # Taking mean of all the points having same cluster label \n",
        "            # To get the mean coordinates find the mean of all the dimensional coordinates\n",
        "            # E.g -> [x,y,z] otherwise mean will be a number and for distance dims must be equal\n",
        "            # experession below will give direct mean instead of coordinates.\n",
        "        # k_centroids.append((np.mean([point for point,cluster in zip(X,y) if cluster == j])))\n",
        "        return k_centroids\n",
        "\n",
        "    def get_cluster(self,X,k_centroids):\n",
        "        y = np.zeros(len(X),dtype=int)\n",
        "        min_dist = None\n",
        "        cluster = 0\n",
        "        dist = None\n",
        "        for i in range(len(X)):\n",
        "            \n",
        "            min_dist = self.euclidean_distance(X[i],k_centroids[0])\n",
        "            cluster = 0\n",
        "            for j in range(len(k_centroids)):\n",
        "                dist = self.euclidean_distance(X[i],k_centroids[j])\n",
        "                if dist < min_dist:\n",
        "                    min_dist = dist\n",
        "                    cluster = j\n",
        "            y[i] = int(cluster)\n",
        "        return y\n",
        "\n",
        "    def clustering(self,X, k):\n",
        "        min_variance = None\n",
        "        variance = None\n",
        "        best_centroids = None\n",
        "        for i in range(self.iterations):\n",
        "            y = np.zeros(len(X),dtype=int)\n",
        "            k_centroids = []\n",
        "            variance = []\n",
        "\n",
        "            # intializes distinct random k points from dataset\n",
        "            while True:\n",
        "                rand_points = np.random.randint(len(X),size=(k))\n",
        "                if len(set(rand_points)) == len(rand_points):\n",
        "                    break\n",
        "            for points in rand_points:\n",
        "                k_centroids.append(X[points])\n",
        "            min_variance = self.get_variance(X,y,k_centroids)\n",
        "            check_count = 0\n",
        "            # Clustering Loop\n",
        "            while True:\n",
        "                check_count += 1\n",
        "                # Prevents the loop from going to infinity in any possible case\n",
        "                if check_count > len(X)*10:\n",
        "                    raise Exception(\"Error in Cluster Loop. Infite Looping. Run again.\") \n",
        "                \n",
        "                # cluster labeling with given centroids\n",
        "                y = self.get_cluster(X,k_centroids) \n",
        "                \n",
        "                # Generates new centroids by taking mean of all labels\n",
        "                new_centroids = self.get_centroids(X,y,k)\n",
        "                if np.array_equal(k_centroids,new_centroids):\n",
        "                    variance = self.get_variance(X,y,new_centroids)\n",
        "                    break\n",
        "                else:\n",
        "                    k_centroids = new_centroids\n",
        "\n",
        "            variance = self.get_variance(X,y,new_centroids)\n",
        "            if variance < min_variance:\n",
        "                min_variance = variance\n",
        "                best_centroids = k_centroids\n",
        "        return best_centroids,min_variance\n",
        "\n",
        "    def elbow_clustering(self,X):\n",
        "        var_reductions = np.zeros(self.k_max+1)\n",
        "        variance = None\n",
        "        variances = []\n",
        "        for i in range(1,int(self.k_max+1)):\n",
        "            _, variance = self.clustering(X,i)\n",
        "            variances.append(variance)\n",
        "        for i in range(len(variances)-1):\n",
        "            var_reductions[i+1] = abs(variances[i+1]-variances[i])\n",
        "        self.elbow_reductions = var_reductions\n",
        "        self.variances = variances\n",
        "        return var_reductions\n",
        "    \n",
        "    def predict(self,X):\n",
        "        y = self.get_cluster(X,self.k_centroids)\n",
        "        return np.array(y).astype(int)\n",
        "\n",
        "    def fit(self, X):\n",
        "        best_k = None\n",
        "        k = self.k\n",
        "        centroids = []\n",
        "        y = []\n",
        "        if self.elbow_flag:\n",
        "\n",
        "            #Gives variances for all the values of K\n",
        "            #implement automation later\n",
        "            self.elbow_clustering(X)\n",
        "\n",
        "        else:\n",
        "            centroids, min_var = self.clustering(X,k) #final centroids are stored globally\n",
        "            y = self.get_cluster(X, centroids)\n",
        "            self.k_centroids = centroids\n",
        "            self.y = y\n",
        "            return y\n",
        "            \n"
      ],
      "execution_count": 10,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Fkfwx-T1co47"
      },
      "source": [
        "class RBFNN():\n",
        "    def __init__(self):\n",
        "        self.k_centroids = None\n",
        "        self.hidden_layer = None\n",
        "        self.weights = None\n",
        "        self.sigma = None   #standard deviation\n",
        "\n",
        "    def euclidean_distance(self,p1,p2):\n",
        "        return np.linalg.norm(p1-p2)\n",
        "\n",
        "    def convert_to_one_hot(self, y, num_of_classes):\n",
        "        arr = np.zeros((len(y), num_of_classes))\n",
        "        for i in range(len(y)):\n",
        "            c = int(y[i])\n",
        "            arr[i][c] = 1\n",
        "        return arr\n",
        "        \n",
        "    def radial_basis_func(self,x,centre,s):\n",
        "        distance = np.linalg.norm(x-centre)\n",
        "        return 1 / np.exp(-distance / s ** 2)\n",
        "\n",
        "    def get_rbf(self,X):\n",
        "        RBF_list = []\n",
        "        for x in X:\n",
        "            RBF_list.append([self.radial_basis_func(x, c, s) for (c, s) in zip(self.k_centroids, self.sigma)])\n",
        "        return np.array(RBF_list)\n",
        "    \n",
        "    def predict(self,X):\n",
        "        RBF_list_tst = self.get_rbf(X)\n",
        "        y_pred = None\n",
        "        y_pred = RBF_list_tst @ self.weights\n",
        "        y_pred = np.array([np.argmax(x) for x in y_pred])\n",
        "        return y_pred\n",
        "\n",
        "    def fit(self,X,Y,classes,centroids,std_deviation):\n",
        "        self.k_centroids = centroids\n",
        "        self.sigma = std_deviation\n",
        "    \n",
        "        self.hidden_layer = self.get_rbf(X)\n",
        "\n",
        "        self.weights = np.linalg.pinv(self.hidden_layer.T @ self.hidden_layer) @ self.hidden_layer.T @ self.convert_to_one_hot(Y, classes)\n",
        "\n",
        "\n"
      ],
      "execution_count": 63,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iYzM7sCtbx28"
      },
      "source": [
        "clf = K_Means(k=3)\n",
        "y_pred = clf.fit(x_train)\n",
        "centroids = clf.k_centroids"
      ],
      "execution_count": 68,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pO7f_MlMchYu",
        "outputId": "397525e6-ad95-4337-8418-55bacded8a85",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "centroids"
      ],
      "execution_count": 69,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[array([-1.08352401,  0.92603438, -1.41251102, -1.34033791]),\n",
              " array([0.96922479, 0.06777268, 0.85939539, 0.88272313]),\n",
              " array([-0.26579651, -0.99255735,  0.20354199,  0.10162204])]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 69
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "idVPBLvzcoTZ"
      },
      "source": [
        "c1,c2,c3 = [],[],[]\n",
        "for x,y in zip(x_train,y_pred):\n",
        "    if y==0:\n",
        "        c1.append(x)\n",
        "    if y==1:\n",
        "        c2.append(x)\n",
        "    else: c3.append(x)\n",
        "standard_deviation = np.array([np.std(c1),np.std(c2),np.std(c3)])"
      ],
      "execution_count": 70,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "QwYdRvGcl98N"
      },
      "source": [
        "rbf = RBFNN()\n",
        "rbf.fit(X= x_train,Y= y_train,classes = 3,centroids = centroids,std_deviation = standard_deviation)"
      ],
      "execution_count": 71,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "i7R5CmpoqLpY"
      },
      "source": [
        "y_pred = rbf.predict(x_test)"
      ],
      "execution_count": 72,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "VEQHolGCrIox",
        "outputId": "0dcc47ec-638e-4e5c-8526-2878f9a0343b",
        "colab": {
          "base_uri": "https://localhost:8080/"
        }
      },
      "source": [
        "matrix = confusion_matrix(y_test, y_pred)\n",
        "class_acc = matrix.diagonal()/matrix.sum(axis=1)\n",
        "print(\"Accuracy Overall \\t\\t%0.2f\"%(accuracy_score(y_test,y_pred)*100))\n",
        "print(\"Accuracy Class Setosa \\t\\t%0.2f\"%(class_acc[0]*100))\n",
        "print(\"Accuracy Class Virginica \\t%0.2f\"%(class_acc[1]*100))\n",
        "print(\"Accuracy Class Virginica \\t%0.2f\"%(class_acc[2]*100))"
      ],
      "execution_count": 80,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Accuracy Overall \t\t70.00\n",
            "Accuracy Class Setosa \t\t100.00\n",
            "Accuracy Class Virginica \t0.00\n",
            "Accuracy Class Virginica \t100.00\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}