scripts/train_variants.py

"""
Train Multiple Model Variants for Milk Spoilage Classification

This script:
1. Loads training data from CSV files
2. Trains 10 RandomForest model variants with different feature subsets
3. Exports all model artifacts (*.joblib, variants_config.json)
"""

import json
import pandas as pd
import numpy as np
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, classification_report
import joblib
import os
from pathlib import Path


# Define all model variants with feature subsets
MODEL_VARIANTS = {
    'baseline': {
        'name': 'Baseline (All Features)',
        'description': 'Uses all 6 microbiological measurements across all time points',
        'features': ['SPC_D7', 'SPC_D14', 'SPC_D21', 'TGN_D7', 'TGN_D14', 'TGN_D21']
    },
    'scenario_1_days14_21': {
        'name': 'Days 14 & 21',
        'description': 'Uses measurements from days 14 and 21 only',
        'features': ['SPC_D14', 'SPC_D21', 'TGN_D14', 'TGN_D21']
    },
    'scenario_2_days7_14': {
        'name': 'Days 7 & 14',
        'description': 'Uses measurements from days 7 and 14 only',
        'features': ['SPC_D7', 'SPC_D14', 'TGN_D7', 'TGN_D14']
    },
    'scenario_3_day21': {
        'name': 'Day 21 Only',
        'description': 'Uses only day 21 measurements',
        'features': ['SPC_D21', 'TGN_D21']
    },
    'scenario_4_day14': {
        'name': 'Day 14 Only',
        'description': 'Uses only day 14 measurements',
        'features': ['SPC_D14', 'TGN_D14']
    },
    'scenario_5_day7': {
        'name': 'Day 7 Only',
        'description': 'Uses only day 7 measurements',
        'features': ['SPC_D7', 'TGN_D7']
    },
    'scenario_6_spc_all': {
        'name': 'SPC Only (All Days)',
        'description': 'Uses only Standard Plate Count measurements across all days',
        'features': ['SPC_D7', 'SPC_D14', 'SPC_D21']
    },
    'scenario_7_tgn_all': {
        'name': 'TGN Only (All Days)',
        'description': 'Uses only Total Gram-Negative measurements across all days',
        'features': ['TGN_D7', 'TGN_D14', 'TGN_D21']
    },
    'scenario_8_spc_7_14': {
        'name': 'SPC Days 7 & 14',
        'description': 'Uses only SPC measurements from days 7 and 14',
        'features': ['SPC_D7', 'SPC_D14']
    },
    'scenario_9_tgn_7_14': {
        'name': 'TGN Days 7 & 14',
        'description': 'Uses only TGN measurements from days 7 and 14',
        'features': ['TGN_D7', 'TGN_D14']
    }
}

# Feature descriptions (constant across all variants)
FEATURE_DESCRIPTIONS = {
    "SPC_D7": "Standard Plate Count at Day 7 (log CFU/mL)",
    "SPC_D14": "Standard Plate Count at Day 14 (log CFU/mL)",
    "SPC_D21": "Standard Plate Count at Day 21 (log CFU/mL)",
    "TGN_D7": "Total Gram-Negative count at Day 7 (log CFU/mL)",
    "TGN_D14": "Total Gram-Negative count at Day 14 (log CFU/mL)",
    "TGN_D21": "Total Gram-Negative count at Day 21 (log CFU/mL)"
}

# Class descriptions (constant across all variants)
CLASS_DESCRIPTIONS = {
    "PPC": "Post-Pasteurization Contamination",
    "no spoilage": "No spoilage detected",
    "spore spoilage": "Spore-forming bacteria spoilage"
}


def load_data():
    """Load training and test data from CSV files."""
    print("Loading data...")
    
    # Adjust path based on where script is run from
    data_dir = Path(__file__).parent.parent / "data"
    if not data_dir.exists():
        # Try alternate path
        data_dir = Path("data")
    
    train_df = pd.read_csv(data_dir / "train_df.csv")
    test_df = pd.read_csv(data_dir / "test_df.csv")
    
    print(f"✓ Loaded {len(train_df)} training samples and {len(test_df)} test samples")
    
    return train_df, test_df


def train_model(X_train, y_train):
    """Train RandomForest model with best hyperparameters from notebook."""
    # Best hyperparameters from GridSearchCV in original notebook
    model = RandomForestClassifier(
        n_estimators=100,
        max_depth=None,
        min_samples_split=5,
        min_samples_leaf=1,
        random_state=42
    )
    
    model.fit(X_train, y_train)
    
    return model


def evaluate_model(model, X_test, y_test):
    """Evaluate model performance on test set."""
    y_pred = model.predict(X_test)
    accuracy = accuracy_score(y_test, y_pred)
    report = classification_report(y_test, y_pred, output_dict=True)
    
    return accuracy, report


def train_all_variants(train_df, test_df, output_dir="model/variants"):
    """Train all model variants and save artifacts."""
    
    # Create output directory
    output_path = Path(output_dir)
    output_path.mkdir(parents=True, exist_ok=True)
    
    target_col = 'spoilagetype'
    variants_metadata = {}
    
    print("\n" + "=" * 70)
    print("Training All Model Variants")
    print("=" * 70)
    
    for variant_id, variant_config in MODEL_VARIANTS.items():
        print(f"\n{variant_id}")
        print(f"  Name: {variant_config['name']}")
        print(f"  Features: {', '.join(variant_config['features'])}")
        
        # Prepare data for this variant
        features = variant_config['features']
        train_set = train_df[features + [target_col]].dropna()
        test_set = test_df[features + [target_col]].dropna()
        
        X_train = train_set[features]
        y_train = train_set[target_col]
        X_test = test_set[features]
        y_test = test_set[target_col]
        
        print(f"  Training samples: {len(X_train)}, Test samples: {len(X_test)}")
        
        # Train model
        model = train_model(X_train, y_train)
        
        # Evaluate
        test_accuracy, test_report = evaluate_model(model, X_test, y_test)
        train_accuracy = accuracy_score(y_train, model.predict(X_train))
        
        print(f"  Train accuracy: {train_accuracy:.4f}")
        print(f"  Test accuracy: {test_accuracy:.4f}")
        
        # Save model
        model_path = output_path / f"{variant_id}.joblib"
        joblib.dump(model, model_path)
        print(f"  ✓ Saved to {model_path}")
        
        # Store metadata
        variants_metadata[variant_id] = {
            'name': variant_config['name'],
            'description': variant_config['description'],
            'features': features,
            'train_accuracy': float(train_accuracy),
            'test_accuracy': float(test_accuracy),
            'n_train_samples': len(X_train),
            'n_test_samples': len(X_test),
            'classes': list(model.classes_),
            'class_metrics': {
                cls: {
                    'precision': float(test_report[cls]['precision']),
                    'recall': float(test_report[cls]['recall']),
                    'f1-score': float(test_report[cls]['f1-score']),
                    'support': int(test_report[cls]['support'])
                }
                for cls in model.classes_
            }
        }
    
    return variants_metadata


def create_variants_config(variants_metadata, output_dir="model/variants"):
    """Create comprehensive config file for all variants."""
    
    config = {
        'model_type': 'RandomForestClassifier',
        'framework': 'sklearn',
        'task': 'classification',
        'hyperparameters': {
            'n_estimators': 100,
            'max_depth': None,
            'min_samples_split': 5,
            'min_samples_leaf': 1,
            'random_state': 42
        },
        'feature_descriptions': FEATURE_DESCRIPTIONS,
        'class_descriptions': CLASS_DESCRIPTIONS,
        'variants': variants_metadata
    }
    
    config_path = Path(output_dir) / "variants_config.json"
    with open(config_path, 'w') as f:
        json.dump(config, f, indent=2)
    
    print(f"\n✓ Variants config saved to {config_path}")
    
    return config


def print_summary(variants_metadata):
    """Print summary of all trained variants."""
    
    print("\n" + "=" * 70)
    print("Training Summary")
    print("=" * 70)
    
    # Sort by test accuracy
    sorted_variants = sorted(
        variants_metadata.items(),
        key=lambda x: x[1]['test_accuracy'],
        reverse=True
    )
    
    print(f"\n{'Rank':<6} {'Variant':<30} {'Test Acc':<12} {'Features'}")
    print("-" * 70)
    
    for rank, (variant_id, metadata) in enumerate(sorted_variants, 1):
        medal = ['🥇', '🥈', '🥉'][rank - 1] if rank <= 3 else '  '
        features_str = ', '.join(metadata['features'][:2]) + (
            '...' if len(metadata['features']) > 2 else ''
        )
        print(f"{medal} {rank:<4} {variant_id:<30} {metadata['test_accuracy']:.4f}       {features_str}")
    
    print("\n" + "=" * 70)


def main():
    """Main function to train all model variants."""
    
    print("=" * 70)
    print("Milk Spoilage Classification - Multi-Variant Training")
    print("=" * 70)
    
    # Load data
    train_df, test_df = load_data()
    
    # Train all variants
    variants_metadata = train_all_variants(train_df, test_df)
    
    # Create config
    create_variants_config(variants_metadata)
    
    # Print summary
    print_summary(variants_metadata)
    
    print("\n✓ All model variants trained successfully!")
    print(f"\nGenerated files:")
    print(f"  - model/variants/*.joblib (10 model files)")
    print(f"  - model/variants/variants_config.json")
    print(f"\nNext step: Update API to load all variants")


if __name__ == "__main__":
    main()