Time series data can be difficult and frustrating to work with, and the various algorithms that generate models can be quite finicky and hard to tune. This is particularly true if you are working with data that has multiple seasonalities. In addition, traditional time series models like SARIMAX have many stringent data requirements like stationarity and equally spaced values. Other time series models like Recurring Neural Networks with Long-Short Term Memory (RNN-LSTM) can be highly complex and difficult to work with if you don’t have a significant level of understanding about neural network architecture.

In 2017, a few researchers at Facebook published a paper called, “Forecasting at Scale” which introduced the open-source project Facebook Prophet, giving quick, powerful, and accessible time-series modeling.