Summary

In this chapter, we learned that we have to produce a chart of time-series data as a first step to see whether our data is a good fit for forecasting. The data has to have autocorrelation (meaning that there is a relationship between past and present values) for us to be able to forecast with it. We can use the Durbin-Watson statistical test to see whether data has autocorrelation or not. The approach of Durbin-Watson is to research the periodicity of errors or residuals to see whether they have predictable behavior. Most products' sales probably do not have periodic behavior, making it difficult for planners to predict optimal inventory management.

In the next chapter, we will learn how to smooth time-series peaks with a Moving Average (MA). We will combine the Centered Moving Average (CMA) with the residuals of linear regression to build a forecast model.