knn-digit-recognition.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import libraries:\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('digits_recognition.csv', header=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0      0\n",
       "1      0\n",
       "2      0\n",
       "3      0\n",
       "4      0\n",
       "      ..\n",
       "780    0\n",
       "781    0\n",
       "782    0\n",
       "783    0\n",
       "784    0\n",
       "Length: 785, dtype: int64"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#check for missing values\n",
    "data.isna().sum()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(3000, 784)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Extract the feature vector and the output label\n",
    "# the feature vector starts from column 1 to the last column\n",
    "X = data.values[:, 1:]  \n",
    "y = data.values[:, 0]\n",
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x360 with 25 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "_, axarr = plt.subplots(5,5,figsize=(5,5))\n",
    "for i in range(5):\n",
    "    for j in range(5):\n",
    "        ## get an int random number from 5000 examples\n",
    "        r = np.random.randint(X.shape[0])\n",
    "        ## get that image from X\n",
    "        XA = X[r].reshape((28, 28)) \n",
    "        ## show in sub-figure (i, j)\n",
    "        axarr[i,j].imshow(XA, cmap='gray')  \n",
    "        axarr[i,j].axis('off') ## turn off axis...\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# scale the feature vector\n",
    "X.min()\n",
    "X.max()\n",
    "X = X/(X.max() - X.min())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(2400, 784)\n",
      "(600, 784)\n",
      "x_train.shape[0]+x_test.shape[0]= 3000\n"
     ]
    }
   ],
   "source": [
    "# split the feature vector and output label into X_train, X_test, y_train, y_test\n",
    "from sklearn.model_selection import train_test_split\n",
    "x_train, x_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state = 47)\n",
    "print(x_train.shape)\n",
    "print(x_test.shape)\n",
    "print('x_train.shape[0]+x_test.shape[0]=', x_train.shape[0] + x_test.shape[0])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "K= 1  test_score= 0.8260516258459351   train_test= 1.0\n",
      "K= 2  test_score= 0.8410234597527478   train_test= 0.9601590922765255\n",
      "K= 3  test_score= 0.8435922431910192   train_test= 0.9386789848024436\n",
      "K= 4  test_score= 0.8510487690295712   train_test= 0.9224833134498631\n",
      "K= 5  test_score= 0.8539849414035221   train_test= 0.9088533356901433\n",
      "K= 6  test_score= 0.8578998379021233   train_test= 0.9021644927779268\n",
      "K= 7  test_score= 0.849411444004715   train_test= 0.8914768846686887\n",
      "K= 8  test_score= 0.843453737562268   train_test= 0.8826804460583971\n",
      "K= 9  test_score= 0.8380954515773675   train_test= 0.8771801349663622\n",
      "K= 10  test_score= 0.8377709214185045   train_test= 0.8709600929654994\n",
      "K= 11  test_score= 0.835787711978485   train_test= 0.8675140119532563\n",
      "max accuracy  0.8578998379021233   train score:  0.9021644927779268    for K= 6\n"
     ]
    }
   ],
   "source": [
    "# Use knn to find best fit\n",
    "from sklearn.neighbors import KNeighborsRegressor\n",
    "test_scores = []\n",
    "train_scores = []\n",
    "K = []\n",
    "for k in range(1, 12):\n",
    "    knn = KNeighborsRegressor(n_neighbors=k)\n",
    "    knn.fit(x_train, y_train)\n",
    "    K.append(k)\n",
    "    test_scores.append(knn.score(x_test, y_test))\n",
    "    train_scores.append(knn.score(x_train, y_train))\n",
    "    print('K=', K[k-1], ' test_score=', test_scores[k-1], '  train_test=', train_scores[k-1])\n",
    "\n",
    "m = max(test_scores)\n",
    "i = test_scores.index(m)\n",
    "print('max accuracy ', m, '  train score: ', train_scores[i], '   for K=', K[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.8429336873552071\n",
      "KNeighborsRegressor(n_neighbors=2)\n",
      "0.8410234597527478\n"
     ]
    }
   ],
   "source": [
    "from sklearn.model_selection import GridSearchCV   # Grid search with cross validation\n",
    "\n",
    "param = {'n_neighbors': list(range(1, 50))}\n",
    "\n",
    "gs = GridSearchCV(KNeighborsRegressor(), param, cv=4)\n",
    "result = gs.fit(x_train, y_train)\n",
    "\n",
    "print(result.best_score_)\n",
    "print(result.best_estimator_)\n",
    "print(gs.score(x_test, y_test))"
   ]
  }
 ],
 "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}