In the previous chapter, exponential smoothing-based forecasting techniques were covered, which is based on the assumption that time series is composed on deterministic and stochastic terms. The random component is zero out with number of observations considered for the forecasting. This assumes that random noise is truly random and follows independent identical distribution. However, this assumption often tends to get violated and smoothing is not sufficient to model the process and set up a forecasting model.
In these scenarios, auto-regressive models can be very useful as these models adjust immediately using the prior lag values by taking advantage of inherent serial correlation between observations. This chapter introduces forecasting concepts using auto-regressive models. The auto-regressive model includes auto-regressive terms or moving average terms...
