config.py

# Output directory for all artifacts
OUTPUT_DIR = "outputs"

# Maximum number of improvement iterations
MAX_ITERATIONS = 3

# Stop if metric exceeds this threshold
TARGET_METRIC_THRESHOLD = 0.95

# Stop if improvement is below this threshold
IMPROVEMENT_THRESHOLD = 0.01

# Random seed for reproducibility
RANDOM_SEED = 42

# Number of parallel jobs (-1 = use all cores)
N_JOBS = -1


# ==============================================================================
# DATA PROCESSING
# ==============================================================================

# Train/test split ratio
TEST_SIZE = 0.2

# Imputation strategies
NUMERICAL_IMPUTATION_STRATEGY = 'median'  # Options: 'mean', 'median', 'most_frequent'
CATEGORICAL_IMPUTATION_STRATEGY = 'most_frequent'

# Feature engineering
APPLY_SCALING = False  # Set True if using distance-based models (SVM, KNN)
SCALING_METHOD = 'standard'  # Options: 'standard', 'minmax', 'robust'


# ==============================================================================
# MODEL SETTINGS
# ==============================================================================

# Classification models to try
CLASSIFICATION_MODELS = {
    'Logistic Regression': {
        'enabled': True,
        'params': {
            'max_iter': 1000,
            'random_state': RANDOM_SEED
        }
    },
    'Random Forest': {
        'enabled': True,
        'params': {
            'n_estimators': 100,
            'random_state': RANDOM_SEED,
            'n_jobs': N_JOBS
        }
    },
    'Extra Trees': {
        'enabled': True,
        'params': {
            'n_estimators': 100,
            'random_state': RANDOM_SEED,
            'n_jobs': N_JOBS
        }
    },
    'Gradient Boosting': {
        'enabled': True,
        'params': {
            'n_estimators': 100,
            'random_state': RANDOM_SEED
        }
    },
    'SVM': {
        'enabled': False,  # Disabled by default (slow on large datasets)
        'params': {
            'kernel': 'rbf',
            'random_state': RANDOM_SEED
        }
    }
}

# Regression models to try
REGRESSION_MODELS = {
    'Linear Regression': {
        'enabled': True,
        'params': {}
    },
    'Ridge Regression': {
        'enabled': True,
        'params': {
            'random_state': RANDOM_SEED
        }
    },
    'Lasso Regression': {
        'enabled': True,
        'params': {
            'random_state': RANDOM_SEED
        }
    },
    'Random Forest': {
        'enabled': True,
        'params': {
            'n_estimators': 100,
            'random_state': RANDOM_SEED,
            'n_jobs': N_JOBS
        }
    },
    'Extra Trees': {
        'enabled': True,
        'params': {
            'n_estimators': 100,
            'random_state': RANDOM_SEED,
            'n_jobs': N_JOBS
        }
    },
    'Gradient Boosting': {
        'enabled': True,
        'params': {
            'n_estimators': 100,
            'random_state': RANDOM_SEED
        }
    }
}


# ==============================================================================
# HYPERPARAMETER OPTIMIZATION
# ==============================================================================

# Enable hyperparameter optimization (requires Optuna)
ENABLE_HYPERPARAMETER_OPTIMIZATION = True

# Number of optimization trials
OPTUNA_N_TRIALS = 30

# Cross-validation folds for optimization
OPTUNA_CV_FOLDS = 3

# Hyperparameter search spaces
HYPERPARAMETER_SEARCH_SPACES = {
    'RandomForest': {
        'n_estimators': (50, 200),
        'max_depth': (3, 20),
        'min_samples_split': (2, 10),
        'min_samples_leaf': (1, 4)
    },
    'ExtraTrees': {
        'n_estimators': (50, 200),
        'max_depth': (3, 20),
        'min_samples_split': (2, 10)
    },
    'GradientBoosting': {
        'n_estimators': (50, 200),
        'learning_rate': (0.01, 0.3),
        'max_depth': (3, 10)
    }
}


# ==============================================================================
# METRICS
# ==============================================================================

# Primary metric for model selection
# Classification: 'accuracy', 'f1', 'roc_auc', 'precision', 'recall'
# Regression: 'r2', 'rmse', 'mae'
PRIMARY_METRIC_CLASSIFICATION = 'accuracy'
PRIMARY_METRIC_REGRESSION = 'r2'

# Additional metrics to compute and report
ADDITIONAL_METRICS_CLASSIFICATION = ['precision', 'recall', 'f1']
ADDITIONAL_METRICS_REGRESSION = ['rmse', 'mae']


# ==============================================================================
# VISUALIZATION
# ==============================================================================

# Figure size for plots
FIGURE_SIZE = (10, 6)

# DPI for saved plots
PLOT_DPI = 150

# Color scheme
PLOT_COLOR = 'steelblue'

# Number of top features to show in importance plot
TOP_N_FEATURES = 15


# ==============================================================================
# REPORTING
# ==============================================================================

# Include code examples in reports
INCLUDE_CODE_EXAMPLES = True

# Include improvement suggestions
INCLUDE_IMPROVEMENT_SUGGESTIONS = True

# Report verbosity (1=minimal, 2=normal, 3=detailed)
REPORT_VERBOSITY = 2


# ==============================================================================
# MEMORY & STRATEGY
# ==============================================================================

# Enable strategy memory (reuse successful strategies)
ENABLE_STRATEGY_MEMORY = True

# Strategy file format
STRATEGY_FILE_FORMAT = 'json'  # Options: 'json', 'yaml'


# ==============================================================================
# ADVANCED SETTINGS
# ==============================================================================

# Show progress bars (requires tqdm)
SHOW_PROGRESS_BARS = False

# Verbose output during training
VERBOSE_TRAINING = True

# Save intermediate models (not just final)
SAVE_INTERMEDIATE_MODELS = False

# Enable feature selection
ENABLE_FEATURE_SELECTION = False
FEATURE_SELECTION_METHOD = 'mutual_info'  # Options: 'mutual_info', 'chi2', 'f_test'

# Handle class imbalance (classification only)
HANDLE_CLASS_IMBALANCE = False
CLASS_IMBALANCE_METHOD = 'oversample'  # Options: 'oversample', 'undersample', 'smote'


# ==============================================================================
# RESOURCE CONSTRAINTS
# ==============================================================================

# Maximum dataset size (rows) for full processing
MAX_DATASET_SIZE = 1_000_000

# Maximum number of features before feature selection is recommended
MAX_FEATURES_WITHOUT_SELECTION = 100

# Maximum training time per model (seconds, 0=unlimited)
MAX_TRAINING_TIME_PER_MODEL = 0

# Memory limit (GB, 0=unlimited)
MEMORY_LIMIT_GB = 0


# ==============================================================================
# VALIDATION
# ==============================================================================

def validate_config():
    """Validate configuration settings"""
    errors = []

    # Check test size
    if not 0 < TEST_SIZE < 1:
        errors.append("TEST_SIZE must be between 0 and 1")

    # Check thresholds
    if not 0 <= TARGET_METRIC_THRESHOLD <= 1:
        errors.append("TARGET_METRIC_THRESHOLD must be between 0 and 1")

    if not 0 < IMPROVEMENT_THRESHOLD < 1:
        errors.append("IMPROVEMENT_THRESHOLD must be between 0 and 1")

    # Check enabled models
    classification_enabled = any(
        m['enabled'] for m in CLASSIFICATION_MODELS.values()
    )
    regression_enabled = any(
        m['enabled'] for m in REGRESSION_MODELS.values()
    )

    if not classification_enabled:
        errors.append("At least one classification model must be enabled")

    if not regression_enabled:
        errors.append("At least one regression model must be enabled")

    # Report errors
    if errors:
        print("⚠️  Configuration Errors:")
        for error in errors:
            print(f"   - {error}")
        return False

    print("✅ Configuration validated successfully")
    return True


if __name__ == "__main__":
    # Test configuration
    print("ML Agent Configuration")
    print("=" * 60)
    print(f"Output Directory: {OUTPUT_DIR}")
    print(f"Random Seed: {RANDOM_SEED}")
    print(f"Test Size: {TEST_SIZE}")
    print(f"Hyperparameter Optimization: {'Enabled' if ENABLE_HYPERPARAMETER_OPTIMIZATION else 'Disabled'}")
    print(f"Optuna Trials: {OPTUNA_N_TRIALS}")
    print(f"Strategy Memory: {'Enabled' if ENABLE_STRATEGY_MEMORY else 'Disabled'}")
    print("=" * 60)

    # Validate
    validate_config()