Forecasting univariate time series data with seasonal ARIMA

In this recipe, you will be introduced to an enhancement to the ARIMA model for handling seasonality, known as the Seasonal Autoregressive Integrated Moving Average or SARIMA. Like an ARIMA(p, d, q), a SARIMA model also requires (p, d, q) to represent non-seasonal orders. Additionally, a SARIMA model requires the orders for the seasonal component, which is denoted as (P, D, Q, s). Combining both components, the model can be written as a SARIMA(p, d, q)(P, D, Q, s). The letters still mean the same, and the letter case indicates which component. For example, the lowercase letters represent the non-seasonal orders, while the uppercase letters represent the seasonal orders. The new parameter, s, is the number of steps per cycle – for example, s=12 for monthly data or s=4 for quarterly data.

In statsmodels, you will use the SARIMAX class to build a SARIMA model.

In this recipe, you will be working with the milk data...