An implementation of KNN based on Numpy and Pandas

knn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from __future__ import division"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = pd.read_csv('./glass_train.txt')\n",
    "X_test = pd.read_csv('./glass_test.txt')\n",
    "\n",
    "y_train = X_train['CLASS']\n",
    "X_train = X_train.drop(['CLASS', 'ID'], axis=1)\n",
    "\n",
    "y_test = X_test['CLASS']\n",
    "X_test = X_test.drop(['CLASS', 'ID'], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train_mean  = X_train.mean()\n",
    "x_train_std  = X_train.std()\n",
    "\n",
    "x_train_max = X_train.max()\n",
    "x_train_min = X_train.min()\n",
    "\n",
    "# X_train = (X_train - x_train_mean) / (x_train_mean - x_train_std)\n",
    "# X_test = (X_test - x_train_mean) / (x_train_mean - x_train_std)\n",
    "\n",
    "X_train = (X_train - x_train_min) / (x_train_max - x_train_min)\n",
    "X_test = (X_test - x_train_min) / (x_train_max - x_train_min)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = pd.get_dummies(y_train).astype(int)\n",
    "y_test = pd.get_dummies(y_test).astype(int)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "k = 1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "from collections import Counter\n",
    "\n",
    "def predict(data, labels, t_instance, k=5):\n",
    "    dist = np.sqrt(np.sum(np.power(t_instance.values - data.values, 2), axis=1))\n",
    "    \n",
    "    top_k = np.argsort(dist)[:k]\n",
    "    top_classes = labels.iloc[top_k]\n",
    "    \n",
    "    unique, count = np.unique(top_classes.values, return_counts=True, axis=0)\n",
    "    inde = np.argmax(count)\n",
    "    \n",
    "    return unique[inde]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "predictions = np.zeros((107, 6), dtype=int)\n",
    "data = X_train.copy()\n",
    "labels = y_train.copy()\n",
    "last = data.shape[0]\n",
    "\n",
    "for index, row in X_test.iterrows():\n",
    "    predicted_class = predict(data, labels, row, k)\n",
    "    predictions[index, :] = predicted_class"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "result = (predictions == y_test.values).all(axis=1)\n",
    "correct = result[result == True]\n",
    "incorrect = result[result == False]\n",
    "accuracy = len(correct)/len(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7476635514018691"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "accuracy"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Sklearn's implementation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.neighbors import KNeighborsClassifier\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "neigh = KNeighborsClassifier(n_neighbors=k)\n",
    "\n",
    "neigh.fit(X_train, y_train)\n",
    "\n",
    "pred_sklearn = neigh.predict(X_test)\n",
    "score = accuracy_score(y_test, pred_sklearn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.7476635514018691"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "score"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python (tf)",
   "language": "python",
   "name": "tf"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}