Global forecasting models for seasonal time series

This recipe shows how to extend a data module to include extra explanatory variables in a TimeSeriesDataSet class and a DataModule class. We’ll use a particular case about seasonal time series.

Getting ready

We load the dataset that we used in the previous recipe:

N_LAGS = 7
HORIZON = 7
from gluonts.dataset.repository.datasets import get_dataset
dataset = get_dataset('nn5_daily_without_missing', regenerate=False)

This dataset contains time series with a daily granularity. Here, we’ll model weekly seasonality using the Fourier series. Unlike what we did in the previous chapter (in the Handling seasonality: seasonal dummies and Fourier series recipe), we’ll learn how to include these features using the TimeSeriesDataSet framework.

How to do it…

Here’s the updated DataModule that includes the Fourier series. We only describe part of the setup() method for brevity. The remaining...