K_means_Class.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "class K_means:\n",
    "\n",
    "    def __init__(self , X , K , Max_iterations):\n",
    "        self.X = X\n",
    "        self.K = K\n",
    "        self.Max_iterations = Max_iterations\n",
    "        self.idx = np.zeros((self.X.shape[0], 1), dtype=np.int8)\n",
    "        self.centroids = np.zeros((K,X.shape[1]))\n",
    "    def kMeansInitCentroids(self):\n",
    "        import random\n",
    "        random_index = random.sample(range(0 , self.X.shape[0]) , self.K)\n",
    "        self.centroids = self.X[random_index]\n",
    "\n",
    "        return self.centroids\n",
    "    \n",
    "    \n",
    "    \n",
    "    def findClosestCentroids(self):\n",
    "        for one_point , i in zip(self.X , range(self.X.shape[0])):\n",
    "            self.idx[i] = np.argmin(np.linalg.norm(one_point-self.centroids , axis = 1))\n",
    "        return self.idx\n",
    "    \n",
    "    def computeCentroids(self):\n",
    "        self.idx = self.idx.ravel()\n",
    "        Cost_function = 0\n",
    "        cost_centroid = []\n",
    "        for i in range(self.K):\n",
    "            self.centroids[i] =  np.mean(self.X[self.idx == i] , axis = 0)\n",
    "            cost_centroid.append(np.array(self.X[self.idx == i]))\n",
    "        for i , j in zip(self.centroids , cost_centroid):\n",
    "            Cost_function += np.sum(np.square(i-j))\n",
    "\n",
    "        return self.centroids , Cost_function\n",
    "\n",
    "    \n",
    "    def k_kmeans_fit(self):\n",
    "        self.centroids = self.kMeansInitCentroids()\n",
    "        cost = 0\n",
    "        for i in range(self.Max_iterations):\n",
    "            self.idx = self.findClosestCentroids()\n",
    "            self.centroids , cost = self.computeCentroids()\n",
    "        return self.centroids , cost ,self.idx\n",
    "    \n",
    "    def predict(self , example):\n",
    "        self.centroids = self.k_kmeans_fit()[0]\n",
    "        prediction = np.argmin(np.linalg.norm(example-self.centroids , axis = 1))\n",
    "        return prediction\n",
    "\n",
    "    \n",
    "    def elbow(self):\n",
    "        import matplotlib.pyplot as plt\n",
    "        K_list = []\n",
    "        cost_list = []\n",
    "        max_slope = 0\n",
    "        prefect_K = 0\n",
    "        for i in range(1,10):\n",
    "            _ , cost ,_ = k_kmeans(self.X , i , self.Max_iterations)\n",
    "            K_list.append(i)\n",
    "            cost_list.append(cost)\n",
    "            if  i-1 >1 and cost_list[i-1]-cost_list[i-2] > max_slope :\n",
    "                max_slope = cost_list[i-1]-cost_list[i-2]\n",
    "                prefect_K = i-1\n",
    "        plt.plot(range(1,10) , cost_list) \n",
    "        plt.xlabel(\"num of K\")\n",
    "        plt.ylabel(\"Cost\")\n",
    "        plt.show()\n",
    "\n",
    "        return prefect_K\n",
    "    \n",
    "    \n"
   ]
  },
  {
   "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": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}