Forecasting time series data using Facebook Prophet

The Prophet library is a popular open-source project that was initially developed at Facebook (now Meta), based on a 2017 paper that proposed an algorithm for time series forecasting titled Forecasting at Scale. The project gained popularity due to its simplicity, ability to create performant forecasting models, and ability to handle complex seasonality, holiday effects, missing data, and outliers. Prophet automates many aspects of designing a forecasting model while providing rich built-in visualizations. Additional capabilities include building growth models (like saturated forecasts), working with uncertainty in trend and seasonality, and detecting changepoints.

In this recipe, you will use the Milk Production dataset for benchmarking performance. This is the same dataset introduced in Chapter 10, Building Univariate Time Series Models Using Statistical Methods. Using the same dataset helps in understanding and comparing different...