Nixtla time series forecasting plugins for Claude Code. StatsForecast, MLForecast, and NeuralForecast integrations with agent skills.
Tier-based daily price-forecasting benchmark for 49 Nifty 50 stocks (1999–2026): baselines vs classical (ARIMA/ETS/Theta/CES) vs global LightGBM, walk-forward CV, leakage-audited, with a full ML thesis. Runs on a Raspberry Pi 5.
Multi-store retail demand forecasting: LightGBM vs classical baselines with rolling-origin backtest
Drop-in forecasting for Prometheus metrics, PromQL in, forecasts out as first-class metrics. Runs statsforecast models against a long-term TSDB and serves predictions on /metrics. Helm-installable, stateless.
Hierarchical demand forecasting · M5 Walmart · 3049 series · 6 levels · MinT reconciliation · LightGBM · DuckDB · conformal prediction · FastAPI · 135 tests
Reproducible demand-forecasting benchmark for GCC retail: Hijri-aware calendar regressors cut WAPE 14% and collapse forecast bias from +8.5% to -0.6%
To associate your repository with the statsforecast topic, visit your repo's landing page and select "manage topics."