
from memorybench79
Run a reproducible univariate time-series forecasting pipeline using StatsForecast models and Polars, producing constrained ensemble forecasts, robust metrics,
This skill runs a full univariate forecasting pipeline using StatsForecast (AutoARIMA, AutoETS, DynamicOptimizedTheta) with Polars for data handling. It prepares data by concatenating ID fields, enforces non-negative constraints on model outputs, builds an ensemble forecast, computes robust group metrics (WMAPE, bias) with outlier-aware filtering, and returns neatly typed and ordered output.
Use when you need production-ready weekly forecasts across many series with strict type handling and when ensemble averaging and non-negative constraints are required (e.g., demand forecasting). Also suitable for cross-validation experiments and benchmark comparisons.
Best for Python-capable agent runtimes that can run StatsForecast and Polars (agent notebooks or server-side Python agents).
This skill has not been reviewed by our automated audit pipeline yet.