kmeans_demo.py

#!/usr/bin/env python3
"""
Simple KMeans demo with PCA for wine quality dataset
"""

import pandas as pd
import numpy as np
import json

def simple_pca(X, n_components=2):
    """Simple PCA implementation"""
    # Center the data
    X_centered = X - np.mean(X, axis=0)
    
    # Compute covariance matrix
    cov_matrix = np.cov(X_centered.T)
    
    # Compute eigenvalues and eigenvectors
    eigenvalues, eigenvectors = np.linalg.eigh(cov_matrix)
    
    # Sort by eigenvalues (descending)
    idx = np.argsort(eigenvalues)[::-1]
    eigenvalues = eigenvalues[idx]
    eigenvectors = eigenvectors[:, idx]
    
    # Select top n_components
    components = eigenvectors[:, :n_components]
    
    # Transform data
    X_pca = X_centered @ components
    
    # Calculate explained variance ratio
    explained_variance_ratio = eigenvalues[:n_components] / np.sum(eigenvalues)
    
    return X_pca, components, explained_variance_ratio

def simple_kmeans(X, k, max_iters=100, random_state=42):
    """Simple KMeans implementation"""
    np.random.seed(random_state)
    
    # Initialize centroids randomly
    n_samples, n_features = X.shape
    centroids = X[np.random.choice(n_samples, k, replace=False)]
    
    for _ in range(max_iters):
        # Assign points to closest centroid
        distances = np.sqrt(((X - centroids[:, np.newaxis])**2).sum(axis=2))
        labels = np.argmin(distances, axis=0)
        
        # Update centroids
        new_centroids = np.array([X[labels == i].mean(axis=0) for i in range(k)])
        
        # Check for convergence
        if np.allclose(centroids, new_centroids):
            break
            
        centroids = new_centroids
    
    # Calculate inertia
    inertia = sum([np.sum((X[labels == i] - centroids[i])**2) for i in range(k)])
    
    return labels, centroids, inertia

def standardize_data(X):
    """Standardize data to have mean=0 and std=1"""
    return (X - np.mean(X, axis=0)) / np.std(X, axis=0)

def analyze_wine_data():
    """Main analysis function"""
    # Load data
    try:
        df = pd.read_csv('data/winequality-merged.csv')
        print(f"✓ Loaded dataset with {len(df)} samples")
    except Exception as e:
        print(f"❌ Error loading data: {e}")
        return None
    
    # Prepare features (exclude wine_type)
    numeric_cols = df.select_dtypes(include=[np.number]).columns.tolist()
    X = df[numeric_cols].values
    
    print(f"✓ Using {len(numeric_cols)} numeric features: {numeric_cols}")
    
    # Standardize data
    X_scaled = standardize_data(X)
    print("✓ Data standardized")
    
    # Apply PCA
    X_pca, components, explained_var = simple_pca(X_scaled, n_components=2)
    print(f"✓ PCA applied - Explained variance: {explained_var[0]:.3f}, {explained_var[1]:.3f}")
    print(f"✓ Total variance explained: {sum(explained_var):.1%}")
    
    # Apply KMeans with different K values
    results = {}
    for k in [2, 3, 4, 5]:
        labels, centroids, inertia = simple_kmeans(X_scaled, k, random_state=42)
        
        # Transform centroids to PCA space
        centroids_centered = centroids - np.mean(X_scaled, axis=0)
        centroids_pca = centroids_centered @ components
        
        results[k] = {
            'labels': labels.tolist(),
            'centroids_pca': centroids_pca.tolist(),
            'inertia': inertia
        }
        
        print(f"✓ KMeans with k={k} - Inertia: {inertia:.2f}")
    
    # Prepare output data
    output_data = {
        'pca_data': X_pca.tolist(),
        'explained_variance': explained_var.tolist(),
        'feature_names': numeric_cols,
        'wine_types': df['wine_type'].tolist() if 'wine_type' in df.columns else None,
        'kmeans_results': results,
        'pca_components': components.tolist()
    }
    
    # Save results
    with open('data/kmeans_results.json', 'w') as f:
        json.dump(output_data, f, indent=2)
    
    print("✓ Results saved to 'data/kmeans_results.json'")
    
    # Print summary
    print("\n" + "="*50)
    print("📊 WINE QUALITY KMEANS ANALYSIS SUMMARY")
    print("="*50)
    print(f"Dataset: {len(df)} samples, {len(numeric_cols)} features")
    if 'wine_type' in df.columns:
        wine_counts = df['wine_type'].value_counts()
        print(f"Wine types: {dict(wine_counts)}")
    
    print(f"\nPCA Results:")
    print(f"  • PC1 explains {explained_var[0]:.1%} of variance")
    print(f"  • PC2 explains {explained_var[1]:.1%} of variance") 
    print(f"  • Total: {sum(explained_var):.1%} of variance captured")
    
    print(f"\nKMeans Results:")
    for k, result in results.items():
        print(f"  • K={k}: Inertia = {result['inertia']:.2f}")
    
    # Feature importance in PCA
    print(f"\nTop 5 features for PC1:")
    feature_importance = [(abs(components[i, 0]), numeric_cols[i], components[i, 0]) 
                         for i in range(len(numeric_cols))]
    feature_importance.sort(reverse=True)
    
    for i, (abs_val, feature, val) in enumerate(feature_importance[:5]):
        print(f"  {i+1}. {feature}: {val:.3f}")
    
    return output_data

if __name__ == "__main__":
    analyze_wine_data()