ARIMA, SARIMA, ETS — Forecasting Pipelines in R
I build production-ready forecasting workflows in R covering ARIMA family models and ETS exponential smoothing. For ARIMA/SARIMA, I use the Box–Jenkins cycle: stationarity checks (ADF/KPSS), seasonal and regular differencing, identification via ACF/PACF, and selection with AICc/BIC. I configure innovations (Gaussian, t), handle ARIMAX/SARIMAX with exogenous regressors, and support automatic model selection alongside manual expert tuning.
For ETS, I cover additive/multiplicative components (A/N, A/A, A/M, M/A, M/M), damped trend, and model comparison with information criteria. I also implement robust preprocessing: calendar/holiday effects, transformations (e.g., Box–Cox), outlier detection (additive, level shift, transient), and missing-value imputation. Forecast evaluation is done with rolling-origin cross-validation and scale-free metrics (MASE, sMAPE, RMSE), with calibrated prediction intervals and multi-horizon accuracy tables.
I’m deeply familiar with these methods and packages (e.g., forecast, fable, feasts, tsibble), and can help you implement end-to-end pipelines, write clear interpretations, and produce publication-quality figures. Deliverables include clean R scripts, diagnostics, and a concise memo you can adapt into reports or theses.
Get help: engagements start at USD $150 with fixed quotes after reviewing your brief and data.
