Book Image

Forecasting Time Series Data with Facebook Prophet

By : Greg Rafferty
Book Image

Forecasting Time Series Data with Facebook Prophet

By: Greg Rafferty

Overview of this book

Prophet enables Python and R developers to build scalable time series forecasts. This book will help you to implement Prophet’s cutting-edge forecasting techniques to model future data with higher accuracy and with very few lines of code. You will begin by exploring the evolution of time series forecasting, from the basic early models to the advanced models of the present day. The book will demonstrate how to install and set up Prophet on your machine and build your first model with only a few lines of code. You'll then cover advanced features such as visualizing your forecasts, adding holidays, seasonality, and trend changepoints, handling outliers, and more, along with understanding why and how to modify each of the default parameters. Later chapters will show you how to optimize more complicated models with hyperparameter tuning and by adding additional regressors to the model. Finally, you'll learn how to run diagnostics to evaluate the performance of your models and see some useful features when running Prophet in production environments. By the end of this Prophet book, you will be able to take a raw time series dataset and build advanced and accurate forecast models with concise, understandable, and repeatable code.
Table of Contents (18 chapters)
1
Section 1: Getting Started
4
Section 2: Seasonality, Tuning, and Advanced Features
13
Section 3: Diagnostics and Evaluation

Performing forward-chaining cross-validation

Forward-chaining cross-validation, also called rolling-origin cross-validation, is similar to k-fold but suited to sequential data such as time series. There is no random shuffling of data to begin but a test set may be set aside. The test set must be the final portion of data, so if each fold is going to be 10% of your data (as it would be in 10-fold cross-validation), then your test set will be the final 10% of your date range.

With the remaining data, you choose an initial amount of data to train on, let's say five folds in this example, and then you evaluate on the sixth fold and save that performance metric. You re-train now on the first six folds and evaluate on the seventh. You repeat until all folds are exhausted and again take the average of your performance metric. The folds using this technique would look like this:

Figure 11.4 – Forward-chaining cross-validation with five folds

In this...