train_model.py

# train_model.py
import numpy as np
from lightgbm import LGBMRegressor
from feature_engineering import extract_and_engineer_features

def run_model_forecasting():
    df = extract_and_engineer_features()
    
    # Select features for machine learning
    features = ['day_of_week', 'month', 'is_weekend', 'sales_lag_7', 'sales_lag_14', 'rolling_mean_7', 'rolling_mean_14']
    target = 'units_sold'
    
    # Split into train set and validation set (most recent 30 days)
    train_df = df.iloc[:-30]
    val_df = df.iloc[-30:]
    
    X_train, y_train = train_df[features], train_df[target]
    X_val, y_val = val_df[features], val_df[target]
    
    # Initialize and train LightGBM model
    model = LGBMRegressor(n_estimators=100, learning_rate=0.05, random_state=42, verbose=-1)
    model.fit(X_train, y_train)
    
    # Evaluate performance accuracy
    predictions = model.predict(X_val)
    mape = np.mean(np.abs((y_val - predictions) / y_val)) * 100
    accuracy = 100 - mape
    print(f"📊 Model Training Performance: Model Accuracy is {accuracy:.2f}%")
    
    # Calculate stable upcoming 14-day demand forecast window
    avg_predicted_daily_sales = model.predict(df[features].tail(14)).mean()
    total_projected_14_day_demand = int(avg_predicted_daily_sales * 14)
    
    print(f"🔮 Forecast Phase Complete: Projected 14-day sales volume is {total_projected_14_day_demand} units.")
    return total_projected_14_day_demand

if __name__ == '__main__':
    run_model_forecasting()