forecasting-ch06-arimalisted

# Chapter 6: The Unexpected Route ## Scope and intake Use when a regular series has lag dependence that may support ARIMA or seasonal ARIMA. Do not use for a multivariate, volatility or causal problem presented as though univariate ARIMA already solves it. Ask only for unresolved material inputs: What is the target frequency and forecast horizon? Are trends deterministic or accumulated shocks? What seasonality exists? Will any future regressors actually be known? ## Input contract and additional evidence CSV timestamp,target, unique regular observations; config horizon,season,as_of. The tool searches seasonal ARIMA orders by AICc after diagnostic differencing (KPSS and seasonal strength), alongside the airline model and the baselines; orders are not fixed in advance. The rolling comparison needs 2 seasons + 4 horizons of history (72 monthly points for a 12-month horizon, 48 for six months, 36 for three); with less, the tool returns a provisional persistence baseline and says how many points are missing. Future exogenous values require their own known-at-origin schedule or explicit scenarios. Minimal **format illustration**, not sufficient training data: ```csv timestamp,target 2026-01-01,100 2026-01-02,102 2026-01-03,101 ``` ## Executable interface Exact CLI columns: `timestamp,target`. Supported method controls: `as_of, conformal, country, criterion, frequency, future_regressors, horizon, origins, per_horizon_buckets, periods, pool, regressors, season, seed, transfo