OptimizeKNN.py

import pandas as pd
import matplotlib.pyplot as plt
from sklearn.neighbors import KNeighborsClassifier  # Updated import for classifier
from sklearn.model_selection import RandomizedSearchCV
from sklearn.metrics import accuracy_score, recall_score, precision_score, f1_score, roc_auc_score
from scipy.stats import randint

def optimize_knn(train_x, train_y,test_x, test_y, fig):
    # Define KNN hyperparameter search space
    param_dist = {
        'n_neighbors': randint(1, 20),
        'weights': ['uniform', 'distance'],
        'p': [1, 2]  # 1 for Manhattan distance (L1), 2 for Euclidean distance (L2)
    }

    # Create KNN model
    knn_model = KNeighborsClassifier()  # Use KNeighborsClassifier

    # Use RandomizedSearchCV for random search
    random_search = RandomizedSearchCV(
        knn_model,
        param_distributions=param_dist,
        n_iter=10,  # Set the number of search iterations
        scoring='accuracy',  # Use accuracy for classification problems
        cv=5,  # Number of cross-validation folds
        random_state=42
    )

    # Perform the search
    random_search.fit(train_x, train_y)

    # Extract the results from the random search
    results_df = pd.DataFrame(random_search.cv_results_)

    # Output the best parameters
    best_params = random_search.best_params_
    print("Evaluate K-Nearest Neighbors (KNN) Model:")
    print("Best Parameters:", best_params)

    # Output the performance of the best model
    best_model = random_search.best_estimator_
    print("Best Model Performance(accuracy):", random_search.best_score_)

    # Make predictions
    predictions = best_model.predict(test_x)

    # Calculate evaluation metrics
    accuracy = accuracy_score(test_y, predictions)

    # For multiclass classification, use average='weighted' for precision, recall, and f1
    precision = precision_score(test_y, predictions, average='weighted', zero_division=0)
    recall = recall_score(test_y, predictions, average='weighted', zero_division=0)
    f1 = f1_score(test_y, predictions, average='weighted', zero_division=0)
    # Calculate AUC
    if best_model.classes_.shape[0] > 2:  # For multiclass classification
        auc = roc_auc_score(pd.get_dummies(test_y), best_model.predict_proba(test_x), multi_class='ovr')
    else:  # For binary classification
        auc = roc_auc_score(test_y, best_model.predict_proba(test_x)[:, 1])


    # Print other metrics
    print(f"Accuracy: {accuracy:.4f}")
    print(f"Precision: {precision:.4f}")
    print(f"Recall: {recall:.4f}")
    print(f"F1 Score: {f1:.4f}")
    print(f"AUC: {auc:.4f}")
    print("")



    if fig:
        # Visualize the relationship between 'n_neighbors' and performance
        plt.figure(figsize=(10, 6))
        plt.scatter(results_df['param_n_neighbors'], results_df['mean_test_score'])
        plt.title('KNN Hyperparameter Tuning')
        plt.xlabel('Number of Neighbors')
        plt.ylabel('Accuracy')
        plt.show()

    return best_model