Exploring ARIMA Models with Interacted Lagged Variables for Forecasting

Including interactions among the explanatory variables in regression models is a common phenomenon. However, including interactions existing among lagged variables in autoregressive models has not been explored so far. In this paper, Autoregressive Integrated Moving Average (ARIMA) model with interactions among the lagged variables is proposed for improving forecast accuracy. The methodology for identifying the interacted lagged variables and including them in the ARIMA model is suggested. Using five different data sets of different types, the paper explores the effect of interacted lagged variables in ARIMA model. The experimental results exhibit that when interactions do actually exist, ARIMA model with interactions improves the forecast accuracy as compared to ARIMA model without interactions.