faro-core 1.0.0

Enterprise-grade multi-SKU time-series forecasting engine

faro-core

Enterprise-grade multi-SKU time-series forecasting engine. Train and compare multiple models (LightGBM, XGBoost, Prophet, ARIMA, ETS, SARIMAX) per SKU/group with automatic feature engineering and walk-forward validation.

Installation

pip install faro-core

Quick Start

from forecasting_core import ForecastEngine

engine = (
    ForecastEngine()
    .load_data("sales.csv")
    .choose_columns(target="sales", date="date", sku="item_id")
    .configure_features(lags=[1, 7, 14], rolling=[7, 14, 28], calendar=True)
    .configure_training(walk_forward=True, wfv_splits=3)
    .configure_forecast(horizon=14)
    .select_models(["lightgbm", "prophet", "ets"])
    .train()
)

print(engine.get_metrics())
forecast = engine.predict(horizon=14)

From Config File

engine = ForecastEngine.from_config("session_config.json")
engine.train()
report = engine.generate_report()

Features

• Multi-model training per SKU: LightGBM, XGBoost, Prophet, ARIMA, ETS, SARIMAX
• Walk-forward validation with configurable splits
• Automatic feature engineering: lags, rolling stats, EWM, calendar features
• Colombia-specific holiday distances (Easter, Christmas)
• Model registry and ensemble support
• Inventory optimization: service level, safety stock
• Data drift monitoring
• Hyperparameter tuning

License

MIT