time-series-forecaster

# Time Series Forecaster ## Overview The Time Series Forecaster skill provides comprehensive capabilities for predicting business metrics over time using classical statistical methods, machine learning, and deep learning approaches. It supports automated model selection, ensemble forecasting, and uncertainty quantification for robust business planning. ## Capabilities - Classical methods (ARIMA, ETS, Theta) - Machine learning methods (XGBoost, LightGBM for time series) - Deep learning methods (Prophet, N-BEATS, Temporal Fusion Transformer) - Ensemble forecasting - Prediction interval generation - Forecast accuracy metrics (MAPE, RMSE, MASE) - Anomaly detection - Seasonality decomposition ## Used By Processes - Predictive Analytics Implementation - KPI Framework Development - Market Sizing and Opportunity Assessment ## Usage ### Data Input ```python # Time series data configuration time_series_data = { "target": "monthly_revenue", "datetime_column": "date", "frequency": "M", # Monthly "data": [ {"date": "2023-01-01", "value": 1000000, "marketing_spend": 50000}, {"date": "2023-02-01", "value": 1050000, "marketing_spend": 55000}, # ... more data ], "exogenous_variables": ["marketing_spend", "economic_index"], "special_events": [ {"date": "2023-11-24", "event": "black_friday", "impact": "positive"}, {"date": "2023-12-25", "event": "christmas", "impact": "mixed"} ] } ``` ### Model Configuration ```...