demand-forecasting-engine

# Demand Forecasting Engine ## Overview The Demand Forecasting Engine provides comprehensive statistical and machine learning-based demand forecasting capabilities. It supports multiple forecasting algorithms with automatic model selection, ensemble averaging, and continuous accuracy tracking to generate reliable demand predictions for supply chain planning. ## Capabilities - **Time Series Forecasting**: ARIMA, exponential smoothing, Holt-Winters methods - **Machine Learning Models**: XGBoost, LSTM neural networks for complex demand patterns - **Causal Factor Integration**: Incorporate promotions, seasonality, trends, and external drivers - **Demand Sensing**: Short-term signal incorporation for near-term forecast adjustment - **Accuracy Metrics**: MAPE, WMAPE, bias calculation and tracking - **Automatic Model Selection**: Best-fit algorithm selection based on data characteristics - **Ensemble Averaging**: Combine multiple model outputs for improved accuracy - **Confidence Intervals**: Generate prediction intervals for uncertainty quantification - **Forecast Value-Add (FVA) Analysis**: Measure contribution of each forecasting step ## Input Schema ```yaml forecast_request: sku_ids: array[string] # SKUs to forecast historical_data: object # Historical demand data forecast_horizon: integer # Periods to forecast granularity: string # daily, weekly, monthly causal_factors: # Optional external factors promo...