In this chapter, the goal was to discuss how important the element of time is in the field of machine learning and analytics, identify the common traps when analyzing the time series, and demonstrate the techniques and methods to work around these traps. We explored both the univariate and bivariate time series analyses for global temperature anomalies and human carbon dioxide emissions. Additionally, we looked at Granger causality to determine if we can say, statistically speaking, that human CO2 emissions cause surface temperature anomalies. While the results—that the temperature change is caused by CO2 emissions—were compelling, they do not seem definitive. However, it does show that Granger causality is an effective tool in investigating causality in machine learning problems. In the next chapter, we will shift gears and take a look at how to apply learning methods to textual data.

Additionally, keep in mind that in time series analysis, we just skimmed the surface. I encourage you to explore other techniques around changepoint detection, decomposition of time series, nonlinear forecasting, and many others. Although not usually considered part of the machine learning toolbox, I believe you will find it an invaluable addition to yours toolbox.