KNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.decomposition import PCA\n",
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "import time"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Preprocessing the Data\n",
    "train = pd.read_csv('train.csv')\n",
    "test = pd.read_csv('testX.csv')\n",
    "\n",
    "# Remove Id and Label columns from the training data\n",
    "train_x = train.drop(['Id', 'Label'], axis = 1)\n",
    "\n",
    "# Take Label column as target variable\n",
    "train_y = train['Label']\n",
    "\n",
    "# Remove Id column from the test data\n",
    "test = test.drop('Id', axis = 1)\n",
    "\n",
    "# Spilt the data into 90% training and 10% validation data\n",
    "X_train, X_val, y_train, y_val = train_test_split(train_x, train_y, test_size=0.1, random_state=42)\n",
    "\n",
    "# Transform the data\n",
    "scalar = StandardScaler(with_std=False)\n",
    "scalar.fit(X_train)\n",
    "train_scaled = scalar.transform(X_train)\n",
    "val_scaled = scalar.transform(X_val)\n",
    "test_scaled = scalar.transform(test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Normalizing the Data\n",
    "from sklearn.metrics import accuracy_score\n",
    "pca = PCA()\n",
    "pca.fit(train_scaled)\n",
    "train_x_transformed  = pca.transform(train_scaled)\n",
    "val_x_transformed = pca.transform(val_scaled)\n",
    "test_transformed = pca.transform(test_scaled)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "83\n",
      "Variance explained by 85 components: 90.2344871977032\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 792x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "variance = pca.explained_variance_ratio_.cumsum()\n",
    "for i in range(0,len(variance)):\n",
    "  if(variance[i] > 0.90):\n",
    "    print(i)\n",
    "    break;\n",
    "\n",
    "print('Variance explained by 85 components:', variance[85]*100)\n",
    "\n",
    "plt.figure(figsize = (11,6))\n",
    "plt.plot(np.arange(0, 784), variance*100)\n",
    "plt.plot(85, variance[85]*100, 'ro', ms = 15)\n",
    "plt.xlabel('number of components', fontsize = 15)\n",
    "plt.ylabel('cumulative explained variance', fontsize = 15)\n",
    "plt.title('Scree Plot', fontsize = 15)\n",
    "plt.grid(True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:  [85.41666666666666, 85.38333333333334, 85.31666666666666, 85.0, 84.75, 84.5, 84.55, 84.38333333333333, 84.28333333333333, 83.88333333333333]\n",
      "Time:  [1.8680987358093262, 1.8769826889038086, 1.9078962802886963, 1.5987629890441895, 1.6126763820648193, 1.6047091484069824, 1.6106939315795898, 1.6465981006622314, 1.611711025238037, 1.8520114421844482]\n"
     ]
    }
   ],
   "source": [
    "# Check for number of neighbour\n",
    "nbr = [10,20,30,40,50,60,70,80,90,100]\n",
    "acc1 = []\n",
    "time1 = []\n",
    "for i in nbr:\n",
    "    neigh = KNeighborsClassifier(n_neighbors=i)\n",
    "    start1 = time.time()\n",
    "    neigh.fit(train_x_transformed[:,:85], y_train)\n",
    "    stop1 = time.time()\n",
    "    y_pred = neigh.predict(val_x_transformed[:,:85])\n",
    "    accuracy = accuracy_score(y_val, y_pred)*100\n",
    "    acc1.append(accuracy)\n",
    "    time1.append(stop1-start1)\n",
    "print(\"Accuracy: \",acc1)\n",
    "print(\"Time: \", time1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:  [82.81666666666668, 80.9, 84.46666666666667, 83.85000000000001, 85.55, 84.91666666666666, 85.71666666666667, 85.56666666666666, 86.0, 85.41666666666666]\n",
      "Time:  [2.012648344039917, 1.7134194374084473, 1.5907480716705322, 1.5857610702514648, 1.5957343578338623, 1.635655164718628, 1.6037392616271973, 1.607729196548462, 1.622661828994751, 1.6017177104949951]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "nbr = [1,2,3,4,5,6,7,8,9,10]\n",
    "acc2 = []\n",
    "time2 = []\n",
    "for i in nbr:\n",
    "    neigh = KNeighborsClassifier(n_neighbors=i)\n",
    "    start2 = time.time()\n",
    "    neigh.fit(train_x_transformed[:,:85], y_train)\n",
    "    stop2 = time.time()\n",
    "    y_pred = neigh.predict(val_x_transformed[:,:85])\n",
    "    accuracy = accuracy_score(y_val, y_pred)*100\n",
    "    acc2.append(accuracy)\n",
    "    time2.append(stop2-start2)\n",
    "print(\"Accuracy: \", acc2)\n",
    "print(\"Time: \", time2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy:  [85.93333333333332, 85.46666666666667, 85.91666666666666, 85.38333333333334, 85.7, 85.31666666666666, 85.6, 85.5, 85.68333333333334, 85.38333333333334]\n",
      "Time:  [1.7293498516082764, 2.0465657711029053, 1.6864640712738037, 1.622661828994751, 1.6865181922912598, 1.5986993312835693, 2.122333288192749, 2.254971504211426, 1.6087324619293213, 1.5927684307098389]\n"
     ]
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "nbr = [11,12,13,14,15,16,17,18,19,20]\n",
    "acc3 = []\n",
    "time3 = []\n",
    "for i in nbr:\n",
    "    neigh = KNeighborsClassifier(n_neighbors=i)\n",
    "    start3 = time.time()\n",
    "    neigh.fit(train_x_transformed[:,:85], y_train)\n",
    "    stop3 = time.time()\n",
    "    y_pred = neigh.predict(val_x_transformed[:,:85])\n",
    "    accuracy = accuracy_score(y_val, y_pred)*100\n",
    "    acc3.append(accuracy)\n",
    "    time3.append(stop3-start3)\n",
    "print(\"Accuracy: \",acc3)\n",
    "print(\"Time: \", time3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy is  86.0\n",
      "Training time is  1.6575927734375\n",
      "Testing time is  11.023545026779175\n"
     ]
    }
   ],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "import time\n",
    "\n",
    "neigh = KNeighborsClassifier(n_neighbors=9)\n",
    "start_train = time.time()\n",
    "neigh.fit(train_x_transformed[:,:85], y_train)\n",
    "stop_train = time.time()\n",
    "\n",
    "start_test = time.time()\n",
    "y_pred = neigh.predict(val_x_transformed[:,:85])\n",
    "stop_test = time.time()\n",
    "\n",
    "acc4 = accuracy_score(y_val, y_pred)*100\n",
    "print(\"Accuracy is \",acc4)\n",
    "print(\"Training time is \", stop_train-start_train)\n",
    "print(\"Testing time is \", stop_test-start_test)"
   ]
  },
  {
   "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}