In this chapter, we covered auto-regressive models such as a MA model to capture serial correlation using error relationship. On similar lines, AR models were covered, which set up the forecasting using the lags as dependent observations. The AR models are good to capture trend information. The ARMA-based approach was also illustrated, which integrates AR and MA models to capture any time-based trends and catastrophic events leading to a lot of error that will take time to correct such as an economy meltdown. All these models assume stationarity; in scenarios where stationarity is not present, a differencing-based model such as ARIMA is proposed, which performs differencing in time series datasets to remove any trend-related components. The forecasting approaches were illustrated with examples using Python's tsa module.
The current chapter focuses on using statistical...
