hw5/src/P2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')\n",
    "\n",
    "sns.reset_defaults()\n",
    "sns.set_theme(rc={'figure.dpi': 72, 'savefig.dpi': 300,\n",
    "              'figure.autolayout': True})\n",
    "sns.set_style('ticks')\n",
    "sns.set_context('paper')\n",
    "\n",
    "np.set_printoptions(precision=4)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Part a\n",
    "\n",
    "# Load the dataset\n",
    "df = pd.read_csv('data/P2/HCV.csv')\n",
    "X = df.iloc[:, 2:]\n",
    "y = df.iloc[:, 1]\n",
    "\n",
    "# Preprocess\n",
    "# Convert categorical to numerical\n",
    "y = y.replace(['0=Blood Donor', '0s=suspect Blood Donor'], 0)\n",
    "y = y.replace('1=Hepatitis', 1)\n",
    "y = y.replace('2=Fibrosis', 2)\n",
    "y = y.replace('3=Cirrhosis', 3)\n",
    "X['Sex'] = X['Sex'].replace('f', 0)\n",
    "X['Sex'] = X['Sex'].replace('m', 1)\n",
    "\n",
    "# Normalize using standard scaler\n",
    "X = X.apply(lambda x: (x - np.mean(x)) / np.std(x))\n",
    "X = X.fillna(0)\n",
    "X, y = X.values, y.values\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Part b\n",
    "\n",
    "def Euclidean(x, X):\n",
    "    \"\"\"Return the Euclidean distances between\n",
    "    an instance x and the dataset X.\"\"\"\n",
    "    return np.sqrt(np.sum((x - X)**2, axis=1))\n",
    "\n",
    "\n",
    "def CityBlock(x, X):\n",
    "    \"\"\"Return the Manhattan distances between\n",
    "    an instance x and the dataset X.\"\"\"\n",
    "    return np.sum(np.abs(x - X), axis=1)\n",
    "\n",
    "\n",
    "def KMeans(X, ClusterCount=2, Iteration=30, ConvergenceRatio=0.01, DistanceMetric=Euclidean):\n",
    "\n",
    "    # Select clusters' centers\n",
    "    centers = []\n",
    "    selected = set()\n",
    "    idx = list(range(X.shape[0]))\n",
    "\n",
    "    # Select the first center randomly.\n",
    "    c1 = np.random.choice(idx)\n",
    "    selected.add(c1)\n",
    "    centers.append(X[c1])\n",
    "\n",
    "    # Select the farthest points as other centers with higher probabilities.\n",
    "    for i in range(ClusterCount - 1):\n",
    "        dists = np.sum([DistanceMetric(c, X) for c in centers], axis=0)\n",
    "        dists /= np.sum(dists)\n",
    "        while True:\n",
    "            ci = np.random.choice(idx, size=1, replace=False, p=dists)[0]\n",
    "            if ci not in selected:\n",
    "                selected.add(ci)\n",
    "                centers.append(X[ci])\n",
    "                break\n",
    "\n",
    "    # Update clusters\n",
    "    for iter in range(Iteration):\n",
    "        # Assign each instance to the nearest cluster.\n",
    "        clusters = [[] for _ in range(ClusterCount)]\n",
    "        for x in X:\n",
    "            dists = DistanceMetric(x, centers)\n",
    "            clusters[np.argmin(dists)].append(x)\n",
    "\n",
    "        # Update clusters' centers\n",
    "        prev_centers = centers\n",
    "        centers = [np.mean(c, axis=0) for c in clusters]\n",
    "\n",
    "        # Stop updating if converged\n",
    "        if np.sum(np.abs(np.array(centers) - np.array(prev_centers))) < ConvergenceRatio:\n",
    "            break\n",
    "\n",
    "    return np.array(centers)\n",
    "\n",
    "\n",
    "def predict(X, centers, DistanceMetric=Euclidean):\n",
    "    pred = np.zeros(X.shape[0], dtype=int)\n",
    "    for i in range(X.shape[0]):\n",
    "        dists = DistanceMetric(X[i], centers)\n",
    "        pred[i] = np.argmin(dists)\n",
    "    return pred\n",
    "\n",
    "\n",
    "def distance_from_center(X, centers, DistanceMetric=Euclidean):\n",
    "    distance = np.zeros(X.shape[0])\n",
    "    for i in range(X.shape[0]):\n",
    "        dists = DistanceMetric(X[i], centers)\n",
    "        distance[i] = np.min(dists)\n",
    "    return distance\n",
    "\n",
    "\n",
    "def value_counts(a):\n",
    "    return dict(zip(*np.unique(a, return_counts=True)))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset status before removing outliers:\n",
      "Number of instances in each cluster (y_pred): {0: 364, 1: 251}\n",
      "Number of instances in each class (y_true): {0: 540, 1: 24, 2: 21, 3: 30}\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 460.8x345.6 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dataset status after removing outliers:\n",
      "Number of instances in each cluster (y_pred): {0: 361, 1: 232}\n",
      "Number of instances in each class (y_true): {0: 536, 1: 22, 2: 20, 3: 15}\n",
      "Proportion of outliers: 0.0358\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 460.8x345.6 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Part c\n",
    "np.random.seed(2)\n",
    "\n",
    "cents = KMeans(X)\n",
    "y_pred = predict(X, cents)\n",
    "y_dist = distance_from_center(X, cents)\n",
    "\n",
    "# Dataset status before removing outliers\n",
    "print('Dataset status before removing outliers:')\n",
    "print('Number of instances in each cluster (y_pred):', value_counts(y_pred))\n",
    "print('Number of instances in each class (y_true):', value_counts(y))\n",
    "ax = sns.histplot(x=y_dist, kde=True, stat='proportion')\n",
    "ax.set(title=\"Distance of Instances From Closest Cluster's Center\", xlabel='Distance')\n",
    "plt.savefig('P2_c1.png')\n",
    "plt.show()\n",
    "\n",
    "# Remove outliers\n",
    "outlier_thr = 7.0\n",
    "outlier_proportion = np.sum(y_dist >= outlier_thr) / len(y)\n",
    "X = X[y_dist < outlier_thr]\n",
    "y = y[y_dist < outlier_thr]\n",
    "y_pred = y_pred[y_dist < outlier_thr]\n",
    "y_dist = y_dist[y_dist < outlier_thr]\n",
    "\n",
    "# Dataset status after removing outliers\n",
    "print('Dataset status after removing outliers:')\n",
    "print('Number of instances in each cluster (y_pred):', value_counts(y_pred))\n",
    "print('Number of instances in each class (y_true):', value_counts(y))\n",
    "print('Proportion of outliers:', round(outlier_proportion, 4))\n",
    "ax = sns.histplot(x=y_dist, kde=True, stat='proportion')\n",
    "ax.set(title=\"Distance of Instances From Closest Cluster's Center\", xlabel='Distance')\n",
    "plt.savefig('P2_c2.png')\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Part d\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(\n",
    "    X, y, train_size=0.8, stratify=y)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Part e\n",
    "\n",
    "def purity_score(y_true, y_pred):\n",
    "    s = 0\n",
    "    for c in np.unique(y_pred):\n",
    "        _, lcnt = np.unique(y_true[y_pred == c], return_counts=True)\n",
    "        s += np.max(lcnt)\n",
    "    return s / len(y_pred)\n",
    "\n",
    "\n",
    "def entropy_score(y_true, y_pred):\n",
    "    s = 0\n",
    "    for c in np.unique(y_pred):\n",
    "        _, lcnt = np.unique(y_true[y_pred == c], return_counts=True)\n",
    "        m_c = np.sum(lcnt)\n",
    "        E_c = -np.sum([(t/m_c) * np.log2(t/m_c) for t in lcnt])\n",
    "        m = len(y_pred)\n",
    "        s += (m_c/m) * E_c\n",
    "    return s\n",
    "\n",
    "\n",
    "def confusion_matrix(y_true, y_pred, labels):\n",
    "    \"\"\"Customized only for clustering task (Not classification)\"\"\"\n",
    "    m = np.zeros((len(labels), len(labels)), dtype=int)\n",
    "    for i in range(m.shape[0]):\n",
    "        for j in range(m.shape[1]):\n",
    "            m[i, j] = np.sum((y_true == i) & (y_pred == labels[j]))\n",
    "    return m\n",
    "\n",
    "\n",
    "def accuracy_score(y_true, y_pred):\n",
    "    \"\"\"Customized only for clustering task (Not classification)\"\"\"\n",
    "    import itertools\n",
    "    permutations = itertools.permutations(np.unique(y_pred))\n",
    "    max_score = 0\n",
    "    for prt in permutations:\n",
    "        cm = confusion_matrix(y_true, y_pred, prt)\n",
    "        score = np.sum(np.diag(cm)) / len(y_pred)\n",
    "        if score > max_score:\n",
    "            max_score = score\n",
    "    return max_score\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for Euclidean:\n",
      "Purity: 0.9076\n",
      "Entropy: 0.4445\n",
      "Accuracy: 0.5546\n",
      "Centers:\n",
      "[[ 0.0919 -0.923  -0.4066 -0.1159 -0.3178 -0.2336 -0.125  -0.4203 -0.0891\n",
      "  -0.2203 -0.2782 -0.2939]\n",
      " [-0.1091  0.7019  0.4509 -0.0143  0.1852 -0.0198 -0.0503  0.402   0.1467\n",
      "   0.0614  0.0099  0.2907]] \n",
      "\n",
      "Results for CityBlock:\n",
      "Purity: 0.9076\n",
      "Entropy: 0.4394\n",
      "Accuracy: 0.521\n",
      "Centers:\n",
      "[[-0.2049  0.6618  0.4981 -0.0261  0.2626  0.0428 -0.029   0.494   0.1584\n",
      "   0.0696  0.0619  0.4267]\n",
      " [ 0.1556 -0.6562 -0.3314 -0.0904 -0.3213 -0.2636 -0.1349 -0.3931 -0.0672\n",
      "  -0.1886 -0.2875 -0.3421]]\n"
     ]
    }
   ],
   "source": [
    "# Part e\n",
    "cents = KMeans(X_train)\n",
    "y_pred = predict(X_test, cents)\n",
    "y_true = y_test.copy()\n",
    "\n",
    "y_true[y_true == 0] = -1\n",
    "y_true[y_true > 0] = 0\n",
    "y_true[y_true == -1] = 1\n",
    "\n",
    "print('Results for Euclidean:')\n",
    "print('Purity:', round(purity_score(y_true, y_pred), 4))\n",
    "print('Entropy:', round(entropy_score(y_true, y_pred), 4))\n",
    "print('Accuracy:', round(accuracy_score(y_true, y_pred), 4))\n",
    "print('Centers:')\n",
    "print(cents, '\\n')\n",
    "\n",
    "\n",
    "cents = KMeans(X_train, DistanceMetric=CityBlock)\n",
    "y_pred = predict(X_test, cents, DistanceMetric=CityBlock)\n",
    "y_true = y_test.copy()\n",
    "\n",
    "y_true[y_true == 0] = -1\n",
    "y_true[y_true > 0] = 0\n",
    "y_true[y_true == -1] = 1\n",
    "\n",
    "print('Results for CityBlock:')\n",
    "print('Purity:', round(purity_score(y_true, y_pred), 4))\n",
    "print('Entropy:', round(entropy_score(y_true, y_pred), 4))\n",
    "print('Accuracy:', round(accuracy_score(y_true, y_pred), 4))\n",
    "print('Centers:')\n",
    "print(cents)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Results for Euclidean:\n",
      "Purity: 0.9076\n",
      "Entropy: 0.5128\n",
      "Accuracy: 0.3529\n",
      "Centers:\n",
      "[[-4.4094e-02 -1.2586e+00 -7.2146e-02 -1.1434e-01 -3.2487e-01 -3.3597e-01\n",
      "  -2.2311e-01 -1.9316e-01  1.3054e-01 -2.2433e-01 -3.1323e-01  3.3222e-04]\n",
      " [ 2.4205e-01  7.3295e-01  3.0070e-01  2.6591e-01  5.8262e-01  3.9429e-02\n",
      "  -7.0801e-02  8.1864e-01  8.8700e-01  1.1703e-01  2.7220e-01  4.7958e-01]\n",
      " [-4.6167e-01  7.9454e-01  4.7569e-01 -1.9734e-01 -8.5691e-02 -5.7668e-02\n",
      "  -4.1034e-02  1.7461e-01 -3.6383e-01  2.3536e-02 -1.4836e-01  1.6439e-01]\n",
      " [ 8.7309e-01  3.4286e-01 -1.0918e+00 -9.6856e-02 -1.2397e-01  1.7492e-01\n",
      "   2.4608e-01 -1.1139e+00 -7.0732e-01 -1.2378e-01 -1.2042e-01 -1.1360e+00]] \n",
      "\n",
      "Results for CityBlock:\n",
      "Purity: 0.9244\n",
      "Entropy: 0.384\n",
      "Accuracy: 0.3866\n",
      "Centers:\n",
      "[[-0.1567  0.7225  0.6246  0.0674  0.2897  0.0329 -0.0359  0.7792  0.7494\n",
      "   0.1099  0.179   0.5976]\n",
      " [-0.2455  0.4322  0.0484 -0.248   2.7481  1.4877  0.0328 -0.3188 -0.6512\n",
      "  -0.1564  0.7943  0.3351]\n",
      " [ 0.0044 -1.2586 -0.1986 -0.0794 -0.3405 -0.3155 -0.2093 -0.2569  0.0862\n",
      "  -0.244  -0.2984 -0.0901]\n",
      " [ 0.0692  0.7945 -0.0068 -0.1047 -0.2173 -0.1612  0.0099 -0.1094 -0.4226\n",
      "   0.0263 -0.2178 -0.2529]]\n"
     ]
    }
   ],
   "source": [
    "# Part f\n",
    "cents = KMeans(X_train, ClusterCount=4)\n",
    "y_pred = predict(X_test, cents)\n",
    "y_true = y_test.copy()\n",
    "\n",
    "print('Results for Euclidean:')\n",
    "print('Purity:', round(purity_score(y_true, y_pred), 4))\n",
    "print('Entropy:', round(entropy_score(y_true, y_pred), 4))\n",
    "print('Accuracy:', round(accuracy_score(y_true, y_pred), 4))\n",
    "print('Centers:')\n",
    "print(cents, '\\n')\n",
    "\n",
    "\n",
    "cents = KMeans(X_train, ClusterCount=4, DistanceMetric=CityBlock)\n",
    "y_pred = predict(X_test, cents, DistanceMetric=CityBlock)\n",
    "y_true = y_test.copy()\n",
    "\n",
    "print('Results for CityBlock:')\n",
    "print('Purity:', round(purity_score(y_true, y_pred), 4))\n",
    "print('Entropy:', round(entropy_score(y_true, y_pred), 4))\n",
    "print('Accuracy:', round(accuracy_score(y_true, y_pred), 4))\n",
    "print('Centers:')\n",
    "print(cents)\n"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3.10.2 64-bit",
   "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": "3.10.2"
  },
  "orig_nbformat": 4,
  "vscode": {
   "interpreter": {
    "hash": "2b324498af64d22b4773901be112d66dec816013b7f64fed368c8550f7daba2d"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}