Time series forecasting with Facebook's Prophet library
Facebook Prophet is a Python library used for forecasting univariate time series with strong support for seasonality and holiday effects. It is especially suitable for time series with frequent changes of trends and is robust enough to handle outliers.
More specifically, the Prophet model is an additive regression model with the following attributes:
- Piecewise linear or logistic growth trend
- Yearly seasonal component modeled with a Fourier series
- Weekly seasonal component modeled with dummy variables
- A user-provided list of holidays
Installation of Prophet is more complicated, since it requires a compiler. The easiest way to install it is by using Anaconda, as follows:
conda install -c conda-forge fbprophet
The accompanying Git repository contains the conda environment set up with Prophet.
The Prophet library requires the input DataFrame to include two columns—ds for date, and...
