Selecting Orders and Extending to SARIMA
Explore how to choose non-seasonal ARIMA orders by interpreting ACF and PACF plots and extend these models to SARIMA for seasonal time series forecasting. Learn to screen candidate models using AIC and BIC criteria and evaluate forecasts on holdout data, understanding the role of seasonal differencing and lag selection in capturing yearly patterns.
We'll cover the following...
Weekly package volume at a delivery hub can rise and fall in a recognizable annual pattern. A non-seasonal ARIMA model may follow the recent level yet miss recurring holiday peaks and quieter weeks. Staffing shifts need a forecast that represents both short-run dependence and the repeating cycle.
As in the previous lesson, the non-seasonal ARIMA order p is the number of autoregressive lags, d is the number of non-seasonal differences, and q is the number of moving-average error lags. When package volume repeats every year, we extend those choices into a seasonal ARIMA model, often called SARIMA.
SARIMA adds a seasonal order