{ "cells": [ { "cell_type": "code", "execution_count": 40, "id": "3ed630fa-e47f-4ab4-a290-ca90caae2168", "metadata": {}, "outputs": [], "source": [ "# Final Project: Applying Machine Learning on Stroke Dataset with Feature Selection\n", "# Presented to: Dr. Hajar Saleh\n", "\n", "# دكتورة هاجر صالح المحترمة،\n", "# هذا المشروع يهدف إلى تطبيق تقنيات تعلم الآلة على مجموعة بيانات السكتة الدماغية خطوة بخطوة.\n", "# الخطوات التي سأتبعها في هذا المشروع هي:\n", "# 1. تحميل مجموعة البيانات والتحقق من هيكلها.\n", "# 2. التعامل مع القيم المفقودة بشكل مناسب لتحضير البيانات لتعلم الآلة.\n", "# 3. تطبيق نموذج Random Forest مع GridSearchCV لإيجاد أفضل المعلمات وتقييم أداء النموذج.\n", "# 4. استخدام طريقتين لاختيار الميزات (Chi-Square وRFE) لتحديد أهم الميزات.\n", "# 5. تطبيق نموذج Random Forest مرة أخرى على الميزات المختارة وتقييم الأداء.\n", "# 6. حفظ الكود ورفعه مع مجموعة البيانات.\n", "\n", "# النتائج المتوقعة:\n", "# - بعد التعامل مع القيم المفقودة، يجب ألا تحتوي البيانات على أي قيم مفقودة.\n", "# - نموذج Random Forest من المتوقع أن يعطي دقة وأداء جيد (Accuracy, Precision, Recall, F1 Score).\n", "# - اختيار الميزات سيساعدنا في تحديد أهم الميزات التي تؤثر على التنبؤ بالسكتة الدماغية.\n", "# - أداء النموذج قد يتحسن أو يظل مشابهًا بعد اختيار الميزات.\n", "# سأبدأ الآن بالخطوة الأولى." ] }, { "cell_type": "code", "execution_count": 21, "id": "a9745f89-bdea-4eaf-8f71-4ea28e1c21da", "metadata": {}, "outputs": [], "source": [ "# Step 1: Import the required libraries\n", "import pandas as pd\n", "import numpy as np\n", "from sklearn.feature_selection import SelectKBest, chi2, RFE\n", "from sklearn.ensemble import RandomForestClassifier\n", "from sklearn.model_selection import train_test_split, GridSearchCV\n", "from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score\n" ] }, { "cell_type": "code", "execution_count": 22, "id": "76711dd0-708c-405a-8bf2-7881c5853df2", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "First 5 rows of the dataset:\n", " id gender age hypertension heart_disease ever_married work_type \\\n", "0 9046 1 67.0 0.0 NaN 1 2.0 \n", "1 51676 0 61.0 0.0 0.0 1 3.0 \n", "2 31112 1 80.0 0.0 NaN 1 2.0 \n", "3 60182 0 49.0 0.0 0.0 1 2.0 \n", "4 1665 0 79.0 NaN 0.0 1 3.0 \n", "\n", " Residence_type avg_glucose_level bmi smoking_status stroke \n", "0 1 NaN 36.600000 1 NaN \n", "1 0 NaN 28.893237 2 NaN \n", "2 0 105.92 32.500000 2 NaN \n", "3 1 NaN 34.400000 3 NaN \n", "4 0 NaN 24.000000 2 NaN \n", "\n", "Columns and data types:\n", "id int64\n", "gender int64\n", "age float64\n", "hypertension float64\n", "heart_disease float64\n", "ever_married int64\n", "work_type float64\n", "Residence_type int64\n", "avg_glucose_level float64\n", "bmi float64\n", "smoking_status int64\n", "stroke float64\n", "dtype: object\n" ] } ], "source": [ "# Step 2: Load the dataset and check its structure\n", "data = pd.read_csv('healthcare-dataset-stroke-data-cleaned.csv')\n", "\n", "# Display the first 5 rows\n", "print(\"First 5 rows of the dataset:\")\n", "print(data.head())\n", "\n", "# Display the columns and data types\n", "print(\"\\nColumns and data types:\")\n", "print(data.dtypes)" ] }, { "cell_type": "code", "execution_count": 23, "id": "2bdf3afb-7480-4e93-9918-c0448c77d14d", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Dropped the 'id' column.\n", "Missing values before handling:\n", "gender 0\n", "age 0\n", "hypertension 498\n", "heart_disease 276\n", "ever_married 0\n", "work_type 657\n", "Residence_type 0\n", "avg_glucose_level 627\n", "bmi 126\n", "smoking_status 0\n", "stroke 249\n", "dtype: int64\n" ] } ], "source": [ "# Step 3: Check for missing values\n", "# I will drop the 'id' column if it exists, then check for missing values in all columns including 'stroke'\n", "if 'id' in data.columns:\n", " data = data.drop('id', axis=1)\n", " print(\"Dropped the 'id' column.\")\n", "else:\n", " print(\"The 'id' column is already dropped or does not exist.\")\n", "\n", "# Check for missing values\n", "print(\"Missing values before handling:\")\n", "print(data.isnull().sum())" ] }, { "cell_type": "code", "execution_count": 30, "id": "9850c7a9-a4bc-42f1-b4f5-d5a1139ec333", "metadata": { "scrolled": true }, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Number of rows before dropping missing values in 'stroke': 4861\n", "Number of rows after dropping missing values in 'stroke': 4861\n", "Missing values in 'stroke' after dropping: 0\n", "Missing values after handling:\n", "gender 0\n", "age 0\n", "hypertension 0\n", "heart_disease 0\n", "ever_married 0\n", "work_type 0\n", "Residence_type 0\n", "avg_glucose_level 0\n", "bmi 0\n", "smoking_status 0\n", "stroke 0\n", "dtype: int64\n" ] } ], "source": [ "# Step 4: Handle missing values\n", "# First, handle missing values in the target column 'stroke'\n", "# I will drop rows where 'stroke' has missing values because it is the target column\n", "print(\"Number of rows before dropping missing values in 'stroke':\", len(data))\n", "data = data.dropna(subset=['stroke'])\n", "print(\"Number of rows after dropping missing values in 'stroke':\", len(data))\n", "\n", "# Double-check missing values in 'stroke' to confirm they are all handled\n", "print(\"Missing values in 'stroke' after dropping:\", data['stroke'].isnull().sum())\n", "if data['stroke'].isnull().sum() > 0:\n", " print(\"Warning: There are still missing values in 'stroke'. This should not happen!\")\n", "\n", "# Now handle missing values in the features\n", "# Fill missing values in numerical columns with the mean\n", "data['avg_glucose_level'] = data['avg_glucose_level'].fillna(data['avg_glucose_level'].mean())\n", "data['bmi'] = data['bmi'].fillna(data['bmi'].mean())\n", "\n", "# Fill missing values in categorical columns with the mode\n", "data['hypertension'] = data['hypertension'].fillna(data['hypertension'].mode()[0])\n", "data['heart_disease'] = data['heart_disease'].fillna(data['heart_disease'].mode()[0])\n", "data['work_type'] = data['work_type'].fillna(data['work_type'].mode()[0])\n", "data['smoking_status'] = data['smoking_status'].fillna(data['smoking_status'].mode()[0])\n", "\n", "# Check for missing values after handling\n", "print(\"Missing values after handling:\")\n", "print(data.isnull().sum())" ] }, { "cell_type": "code", "execution_count": 31, "id": "e940db1e-4cf4-421b-9bc9-92fffe4ff140", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Checking for missing values in y_train:\n", "0\n", "Checking for missing values in y_test:\n", "0\n", "Shape of X_train: (3888, 10)\n", "Shape of X_test: (973, 10)\n", "Shape of y_train: (3888,)\n", "Shape of y_test: (973,)\n" ] } ], "source": [ "# Step 5: Split the data into features and target, then into training and testing sets\n", "# Separate features (X) and target (y)\n", "X = data.drop('stroke', axis=1)\n", "y = data['stroke']\n", "\n", "# Split the data into training (80%) and testing (20%) sets\n", "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)\n", "\n", "# Check for missing values in y_train and y_test\n", "print(\"Checking for missing values in y_train:\")\n", "print(y_train.isnull().sum())\n", "print(\"Checking for missing values in y_test:\")\n", "print(y_test.isnull().sum())\n", "\n", "# Display the shapes of the training and testing sets\n", "print(\"Shape of X_train:\", X_train.shape)\n", "print(\"Shape of X_test:\", X_test.shape)\n", "print(\"Shape of y_train:\", y_train.shape)\n", "print(\"Shape of y_test:\", y_test.shape)" ] }, { "cell_type": "code", "execution_count": 38, "id": "3ee179f6-d0fd-4879-9b88-731d4359430a", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Evaluation Results (Before Feature Selection):\n", "Model Name: RandomForestClassifier (Before Feature Selection)\n", "Accuracy: 1.0000\n", "Precision: 1.0000\n", "Recall: 1.0000\n", "F1 Score: 1.0000\n", "\n", "Best hyperparameters found by GridSearchCV (Before Feature Selection):\n", "{'max_depth': 10, 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 100}\n" ] } ], "source": [ "# Step 6: Apply Random Forest with GridSearchCV Before Feature Selection\n", "# Initialize the Random Forest Classifier\n", "rf_classifier = RandomForestClassifier(random_state=42)\n", "\n", "# Define the parameter grid for GridSearchCV\n", "param_grid = {\n", " 'n_estimators': [100, 200],\n", " 'max_depth': [10, 20, None],\n", " 'min_samples_split': [2, 5],\n", " 'min_samples_leaf': [1, 2]\n", "}\n", "\n", "# Initialize GridSearchCV with 5-fold cross-validation\n", "grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, cv=5, scoring='accuracy', n_jobs=-1)\n", "\n", "# Fit the grid search to the data\n", "grid_search.fit(X_train, y_train)\n", "\n", "# Get the best estimator from grid search\n", "best_rf_classifier = grid_search.best_estimator_\n", "\n", "# Make predictions using the best model\n", "y_pred = best_rf_classifier.predict(X_test)\n", "\n", "# Evaluate the model\n", "evaluation_results = evaluate_model('RandomForestClassifier (Before Feature Selection)', y_test, y_pred)\n", "\n", "# Print the evaluation results\n", "print(\"Evaluation Results (Before Feature Selection):\")\n", "for key, value in evaluation_results.items():\n", " print(f\"{key}: {value:.4f}\" if isinstance(value, float) else f\"{key}: {value}\")\n", "\n", "# Print the best parameters found by grid search\n", "print(\"\\nBest hyperparameters found by GridSearchCV (Before Feature Selection):\")\n", "print(grid_search.best_params_)" ] }, { "cell_type": "code", "execution_count": 33, "id": "abe017bb-db7c-4788-8e44-b7ccf89c487b", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Checking for missing values in X_train before Chi-Square:\n", "gender 0\n", "age 0\n", "hypertension 0\n", "heart_disease 0\n", "ever_married 0\n", "work_type 0\n", "Residence_type 0\n", "avg_glucose_level 0\n", "bmi 0\n", "smoking_status 0\n", "dtype: int64\n", "\n", "Checking for missing values in X_test before Chi-Square:\n", "gender 0\n", "age 0\n", "hypertension 0\n", "heart_disease 0\n", "ever_married 0\n", "work_type 0\n", "Residence_type 0\n", "avg_glucose_level 0\n", "bmi 0\n", "smoking_status 0\n", "dtype: int64\n", "\n", "Checking for missing values in X_train_non_negative after clipping:\n", "gender 0\n", "age 0\n", "hypertension 0\n", "heart_disease 0\n", "ever_married 0\n", "work_type 0\n", "Residence_type 0\n", "avg_glucose_level 0\n", "bmi 0\n", "smoking_status 0\n", "dtype: int64\n", "\n", "Checking for missing values in X_test_non_negative after clipping:\n", "gender 0\n", "age 0\n", "hypertension 0\n", "heart_disease 0\n", "ever_married 0\n", "work_type 0\n", "Residence_type 0\n", "avg_glucose_level 0\n", "bmi 0\n", "smoking_status 0\n", "dtype: int64\n", "\n", "Feature Scores (Chi-Square Test):\n", " Feature Score\n", "0 gender NaN\n", "1 age NaN\n", "2 hypertension NaN\n", "3 heart_disease NaN\n", "4 ever_married NaN\n", "5 work_type NaN\n", "6 Residence_type NaN\n", "7 avg_glucose_level NaN\n", "8 bmi NaN\n", "9 smoking_status NaN\n" ] } ], "source": [ "# Step 7: Feature Selection - Method 1 (Chi-Square Test)\n", "# Select the top 5 features using chi2\n", "# First, I will double-check for any missing values in X_train and X_test\n", "print(\"Checking for missing values in X_train before Chi-Square:\")\n", "print(X_train.isnull().sum())\n", "print(\"\\nChecking for missing values in X_test before Chi-Square:\")\n", "print(X_test.isnull().sum())\n", "\n", "# If there are any missing values, I will handle them again\n", "X_train = X_train.fillna(X_train.mean())\n", "X_test = X_test.fillna(X_test.mean())\n", "\n", "# Now, I will ensure all values are non-negative for Chi-Square\n", "X_train_non_negative = X_train.copy()\n", "X_test_non_negative = X_test.copy()\n", "for col in X_train_non_negative.columns:\n", " X_train_non_negative[col] = X_train_non_negative[col].clip(lower=0)\n", " X_test_non_negative[col] = X_test_non_negative[col].clip(lower=0)\n", "\n", "# Double-check again to make sure there are no NaN values after clipping\n", "print(\"\\nChecking for missing values in X_train_non_negative after clipping:\")\n", "print(X_train_non_negative.isnull().sum())\n", "print(\"\\nChecking for missing values in X_test_non_negative after clipping:\")\n", "print(X_test_non_negative.isnull().sum())\n", "\n", "# Apply SelectKBest with chi2 to select the top 5 features\n", "select_feature = SelectKBest(chi2, k=5).fit(X_train_non_negative, y_train)\n", "\n", "# Create a DataFrame to store feature scores\n", "feature_scores = pd.DataFrame({'Feature': X.columns, 'Score': select_feature.scores_})\n", "print(\"\\nFeature Scores (Chi-Square Test):\")\n", "print(feature_scores)\n", "\n", "# Transform the training and testing sets with selected features\n", "X_train_selected = select_feature.transform(X_train_non_negative)\n", "X_test_selected = select_feature.transform(X_test_non_negative)" ] }, { "cell_type": "code", "execution_count": 34, "id": "7d13ca57-b95c-4d74-867c-3b1f39bee22c", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Evaluation Results (Chi-Square Selected Features):\n", "Model Name: RandomForestClassifier (Chi-Square)\n", "Accuracy: 1.0000\n", "Precision: 1.0000\n", "Recall: 1.0000\n", "F1 Score: 1.0000\n", "\n", "Best hyperparameters found by GridSearchCV (Chi-Square):\n", "{'max_depth': None, 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 100}\n" ] } ], "source": [ "# Step 8: Apply Random Forest on selected features (Chi-Square)\n", "# Initialize GridSearchCV with the same parameters\n", "grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, cv=5, scoring='accuracy', n_jobs=-1)\n", "\n", "# Fit the grid search to the selected features\n", "grid_search.fit(X_train_selected, y_train)\n", "\n", "# Get the best estimator from grid search\n", "best_rf_classifier = grid_search.best_estimator_\n", "\n", "# Make predictions using the best model\n", "y_pred = best_rf_classifier.predict(X_test_selected)\n", "\n", "# Evaluate the model\n", "evaluation_results = evaluate_model('RandomForestClassifier (Chi-Square)', y_test, y_pred)\n", "\n", "# Print the evaluation results\n", "print(\"Evaluation Results (Chi-Square Selected Features):\")\n", "for key, value in evaluation_results.items():\n", " print(f\"{key}: {value:.4f}\" if isinstance(value, float) else f\"{key}: {value}\")\n", "\n", "# Print the best parameters found by grid search\n", "print(\"\\nBest hyperparameters found by GridSearchCV (Chi-Square):\")\n", "print(grid_search.best_params_)" ] }, { "cell_type": "code", "execution_count": 35, "id": "857d4eb6-722d-4ca5-96cd-6e21f2f13b9b", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "\n", "Selected Features (RFE): [5 6 7 8 9]\n" ] } ], "source": [ "# Step 9: Feature Selection - Method 2 (Recursive Feature Elimination - RFE)\n", "# Use RFE to select the top 5 features\n", "rfe_selector = RFE(estimator=RandomForestClassifier(random_state=42), n_features_to_select=5, step=1)\n", "\n", "# Fit RFE to the training data\n", "X_train_selected_rfe = rfe_selector.fit_transform(X_train, y_train)\n", "X_test_selected_rfe = rfe_selector.transform(X_test)\n", "\n", "# Get the selected feature indices\n", "selected_features = rfe_selector.get_support(indices=True)\n", "print(\"\\nSelected Features (RFE):\", selected_features)" ] }, { "cell_type": "code", "execution_count": 36, "id": "bf49957f-56fc-4a8e-ab0c-b2b106e34fae", "metadata": {}, "outputs": [ { "name": "stdout", "output_type": "stream", "text": [ "Evaluation Results (RFE Selected Features):\n", "Model Name: RandomForestClassifier (RFE)\n", "Accuracy: 1.0000\n", "Precision: 1.0000\n", "Recall: 1.0000\n", "F1 Score: 1.0000\n", "\n", "Best hyperparameters found by GridSearchCV (RFE):\n", "{'max_depth': None, 'min_samples_leaf': 1, 'min_samples_split': 2, 'n_estimators': 100}\n" ] } ], "source": [ "# Step 10: Apply Random Forest on selected features (RFE)\n", "# Initialize GridSearchCV with the same parameters\n", "grid_search = GridSearchCV(estimator=rf_classifier, param_grid=param_grid, cv=5, scoring='accuracy', n_jobs=-1)\n", "\n", "# Fit the grid search to the selected features\n", "grid_search.fit(X_train_selected_rfe, y_train)\n", "\n", "# Get the best estimator from grid search\n", "best_rf_classifier = grid_search.best_estimator_\n", "\n", "# Make predictions using the best model\n", "y_pred = best_rf_classifier.predict(X_test_selected_rfe)\n", "\n", "# Evaluate the model\n", "evaluation_results = evaluate_model('RandomForestClassifier (RFE)', y_test, y_pred)\n", "\n", "# Print the evaluation results\n", "print(\"Evaluation Results (RFE Selected Features):\")\n", "for key, value in evaluation_results.items():\n", " print(f\"{key}: {value:.4f}\" if isinstance(value, float) else f\"{key}: {value}\")\n", "\n", "# Print the best parameters found by grid search\n", "print(\"\\nBest hyperparameters found by GridSearchCV (RFE):\")\n", "print(grid_search.best_params_)" ] }, { "cell_type": "code", "execution_count": 39, "id": "5d016bee-47ca-42dc-bfaf-0317212734f0", "metadata": {}, "outputs": [], "source": [ "# # الخاتمة وحفظ الكود (Conclusion and Save the Code) \n", "# دكتورة هاجر صالح المحترمة، \n", "# لقد أكملت جميع خطوات المشروع: \n", "# - تعاملت مع القيم المفقودة (Missing Values) بملء الأعمدة العددية (Numerical Columns) بالمتوسط (Mean) والأعمدة التصنيفية (Categorical Columns) بالقيمة الأكثر تكرارًا (Mode). \n", "# - قمت بحذف الصفوف (Rows) التي تحتوي على قيم مفقودة في العمود الهدف 'stroke' (Target Column 'stroke') لتجنب التحيز (Bias). \n", "# - طبقت نموذج Random Forest (Random Forest Model) مع GridSearchCV (GridSearchCV) قبل وبعد اختيار الميزات (Feature Selection). \n", "# - استخدمت طريقتين لاختيار الميزات (Feature Selection Methods): (Chi-Square وRFE) لاختيار أفضل 5 ميزات (Top 5 Features). \n", "# - قيّمت أداء النموذج (Model Performance) في جميع الحالات باستخدام الدقة (Accuracy)، الدقة الموجبة (Precision)، الاستدعاء (Recall)، ومقياس F1 (F1 Score). \n", "# النتائج تُظهر أداء النموذج (Model Performance) مع وبدون اختيار الميزات (With and Without Feature Selection). \n", "# سأحفظ هذا الكود باسم 'Hassan_Alzahrani.py' وأرفعه مع مجموعة البيانات (Dataset) 'healthcare-dataset-stroke-data-cleaned.csv'. \n", "# شكرًا لمراجعتكِ لمشروعي!" ] }, { "cell_type": "code", "execution_count": null, "id": "f9e5d8b9-e67e-41ed-b29c-8713e18b67b9", "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": "3.12.7" } }, "nbformat": 4, "nbformat_minor": 5 }