Summary

Seasonality truly is the heart of Prophet. This chapter covered a lot of ground; the foundations you learned here will be used throughout the remaining chapters of this book. Indeed, almost any model you build in Prophet will have seasonality considerations, whereas many of the upcoming chapters cover special cases that may or may not apply to your specific problem.

You started this chapter by learning the difference between additive and multiplicative seasonality, and how to identify whether your dataset features one or the other. We then briefly discussed the Fourier series and demonstrated how a partial Fourier sum can build a very complex periodic curve. Using these ideas, you learned how setting the Fourier order of a seasonality can be used to control its shape by allowing either more or less freedom to bend along its path.

Next, you modeled the 11-year cycle of sunspots and learned how to add custom seasonalities. These custom seasonalities came into use again...