ACOfeatureselection.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "96b07754-725a-406c-b96f-3dba7b2b7297",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Validation Accuracy: 0.984375\n",
      "Validation Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "     defects       1.00      0.97      0.98        29\n",
      "      normal       0.97      1.00      0.99        35\n",
      "\n",
      "    accuracy                           0.98        64\n",
      "   macro avg       0.99      0.98      0.98        64\n",
      "weighted avg       0.98      0.98      0.98        64\n",
      "\n",
      "Test Accuracy: 0.671875\n",
      "Test Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "     defects       0.62      0.72      0.67        29\n",
      "      normal       0.73      0.63      0.68        35\n",
      "\n",
      "    accuracy                           0.67        64\n",
      "   macro avg       0.68      0.68      0.67        64\n",
      "weighted avg       0.68      0.67      0.67        64\n",
      "\n"
     ]
    }
   ],
   "source": [
    "\n",
    "#import libraries\n",
    "\n",
    "import numpy as np\n",
    "import os\n",
    "import cv2\n",
    "import pandas as pd\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import accuracy_score, classification_report\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "\n",
    "#loading datasets and labels\n",
    "\n",
    "dire = \"D:/dataset trail\"\n",
    "categories = ['defects', 'normal']\n",
    "\n",
    "images = []\n",
    "labels = []\n",
    "\n",
    "SIZE = 200\n",
    "\n",
    "for category in categories:\n",
    "    path = os.path.join(dire, category)\n",
    "    label = categories.index(category)\n",
    "    \n",
    "    for img in os.listdir(path):\n",
    "        imgpath = os.path.join(path, img)\n",
    "        tree_img = cv2.imread(imgpath, 0)\n",
    "        try:\n",
    "            tree_img = cv2.resize(tree_img, (SIZE, SIZE))\n",
    "            images.append(tree_img)\n",
    "            labels.append(label)\n",
    "        except Exception as e:\n",
    "            pass\n",
    "\n",
    "images = np.array(images)\n",
    "labels = np.array(labels)\n",
    "\n",
    "\n",
    "#feature extraction using GLCM (Use other better feature extractors for good results)\n",
    "\n",
    "def feature_extractor(dataset):\n",
    "    image_dataset = pd.DataFrame()\n",
    "    for img in dataset:\n",
    "        glcm = cv2.calcHist([img.astype(np.uint8)], [0], None, [256], [0, 256])\n",
    "        contrast = np.sum((np.arange(glcm.shape[0]) - np.arange(glcm.shape[1])) ** 2 * glcm)\n",
    "        dissimilarity = np.sum(np.abs(np.arange(glcm.shape[0]) - np.arange(glcm.shape[1])) * glcm)\n",
    "        homogeneity = np.sum(glcm / (1 + np.abs(np.arange(glcm.shape[0]) - np.arange(glcm.shape[1]))))\n",
    "        energy = np.sum(glcm ** 2)\n",
    "        \n",
    "        df = pd.DataFrame({\n",
    "            'Energy': energy,\n",
    "            'Diss_sim': dissimilarity,\n",
    "            'Homogen': homogeneity,\n",
    "            'Contrast': contrast\n",
    "        }, index=[0])\n",
    "        \n",
    "        image_dataset = pd.concat([image_dataset, df], ignore_index=True)\n",
    "        \n",
    "    return image_dataset\n",
    "\n",
    "\n",
    "X_for_ML = feature_extractor(images)\n",
    "\n",
    "\n",
    "# Train-validation-test split (60-20-20)\n",
    "x_train_val, x_test, y_train_val, y_test = train_test_split(X_for_ML, labels, test_size=0.2, random_state=42)\n",
    "x_train, x_val, y_train, y_val = train_test_split(x_train_val, y_train_val, test_size=0.25, random_state=42)\n",
    "\n",
    "# Feature scaling\n",
    "scaler = StandardScaler()\n",
    "x_train_scaled = scaler.fit_transform(x_train)\n",
    "x_val_scaled = scaler.transform(x_val)\n",
    "x_test_scaled = scaler.transform(x_test)\n",
    "\n",
    "\n",
    "#Random forest declartion\n",
    "def fitness_function(feature_mask):\n",
    "    X_train_subset = x_train_scaled[:, feature_mask]\n",
    "    X_val_subset = x_val_scaled[:, feature_mask]\n",
    "    \n",
    "    clf = RandomForestClassifier()\n",
    "    clf.fit(X_train_subset, y_train)\n",
    "    return clf.score(X_val_subset, y_val)\n",
    "\n",
    "\n",
    "#Aco \n",
    "class ACO:\n",
    "    def __init__(self, num_ants, num_features, alpha, beta, evaporation_rate, q0):\n",
    "        self.num_ants = num_ants\n",
    "        self.num_features = num_features\n",
    "        self.alpha = alpha\n",
    "        self.beta = beta\n",
    "        self.evaporation_rate = evaporation_rate\n",
    "        self.q0 = q0\n",
    "        self.pheromone_matrix = np.ones((num_features, 2)) / num_features\n",
    "\n",
    "    def run(self, num_iterations):\n",
    "        best_feature_mask = None\n",
    "        best_fitness = 0\n",
    "        for iteration in range(num_iterations):\n",
    "            ant_fitnesses = []\n",
    "            for ant in range(self.num_ants):\n",
    "                feature_mask = self.select_features()\n",
    "                ant_fitness = fitness_function(feature_mask)\n",
    "                ant_fitnesses.append((feature_mask, ant_fitness))\n",
    "                if ant_fitness > best_fitness:\n",
    "                    best_fitness = ant_fitness\n",
    "                    best_feature_mask = feature_mask\n",
    "            self.update_pheromone(ant_fitnesses)\n",
    "        return best_feature_mask\n",
    "\n",
    "    def select_features(self):\n",
    "        feature_mask = np.zeros(self.num_features, dtype=bool)\n",
    "        while not feature_mask.any():\n",
    "            q = np.random.uniform()\n",
    "            if q <= self.q0:\n",
    "                feature_index = self.select_best_feature()\n",
    "            else:\n",
    "                feature_index = self.select_random_feature()\n",
    "            feature_mask[feature_index] = True\n",
    "        return feature_mask\n",
    "\n",
    "    def select_best_feature(self):\n",
    "        heuristic_info = self.pheromone_matrix[:, 0] ** self.alpha * (1 / self.pheromone_matrix[:, 1]) ** self.beta\n",
    "        return np.argmax(heuristic_info)\n",
    "\n",
    "    def select_random_feature(self):\n",
    "        return np.random.randint(self.num_features)\n",
    "\n",
    "    def update_pheromone(self, ant_fitnesses):\n",
    "        self.pheromone_matrix *= self.evaporation_rate\n",
    "        for feature_index in range(self.num_features):\n",
    "            for ant_fitness in ant_fitnesses:\n",
    "                if ant_fitness[0][feature_index]:\n",
    "                    self.pheromone_matrix[feature_index, 0] += 1 / ant_fitness[1]\n",
    "\n",
    "# Parameters for the ACO algorithm\n",
    "num_ants = 20\n",
    "num_features = x_train_scaled.shape[1]\n",
    "alpha = 1\n",
    "beta = 1\n",
    "evaporation_rate = 0.3\n",
    "q0 = 0.9\n",
    "num_iterations = 100\n",
    "\n",
    "# Run ACO algorithm\n",
    "aco = ACO(num_ants, num_features, alpha, beta, evaporation_rate, q0)\n",
    "best_feature_mask = aco.run(num_iterations)\n",
    "\n",
    "# Evaluate performance of selected features on validation set\n",
    "X_val_selected = x_val_scaled[:, best_feature_mask]\n",
    "clf = RandomForestClassifier()\n",
    "clf.fit(X_val_selected, y_val)\n",
    "y_val_pred = clf.predict(X_val_selected)\n",
    "val_accuracy = accuracy_score(y_val, y_val_pred)\n",
    "val_classification_report = classification_report(y_val, y_val_pred, target_names=categories)\n",
    "\n",
    "# Train SVM model on selected features using training + validation set\n",
    "X_train_val_selected = np.concatenate((x_train_scaled[:, best_feature_mask], X_val_selected), axis=0)\n",
    "y_train_val_combined = np.concatenate((y_train, y_val), axis=0)\n",
    "clf.fit(X_train_val_selected, y_train_val_combined)\n",
    "\n",
    "# Evaluate performance on testing set\n",
    "X_test_selected = x_test_scaled[:, best_feature_mask]\n",
    "y_test_pred = clf.predict(X_test_selected)\n",
    "test_accuracy = accuracy_score(y_test, y_test_pred)\n",
    "test_classification_report = classification_report(y_test, y_test_pred, target_names=categories)\n",
    "\n",
    "# Print accuracies and classification reports\n",
    "print(\"Validation Accuracy:\", val_accuracy)\n",
    "print(\"Validation Classification Report:\")\n",
    "print(val_classification_report)\n",
    "\n",
    "print(\"Test Accuracy:\", test_accuracy)\n",
    "print(\"Test Classification Report:\")\n",
    "print(test_classification_report)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "48fd6522-f6fe-41b3-a4a0-64a9de5107ad",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}