Time series are everywhere in science and business, and the ability to forecast them accurately and efficiently can provide decisive advantages. Darts is an open-source Python library, which provides a wide variety of forecasting models and tools under a single and user-friendly API. It puts emphasis on reducing the experiment cycle duration and improving the ease of using, comparing and combining different models; from ARIMA to deep learning models.

This talk will give a tour of Darts and some of its main features, such as: quick creation and comparison of forecasting models, backtesting, ML-based models applied to time series forecasting, training forecasting models on multiple time series, producing probabilistic forecasts and integrating external data. We will go over a few toy examples, and see how to address them in a few lines of code.

Goals of the talk:

• Introduce how one can tackle forecasting problems
• Obtain great results quickly in few line of codes

Pre-requisites:

• Basic knowledge of Python
• Basic knowledge of data science & machine learning

Key take-aways:

• Quickly create forecasts with your own data
• Compare and select the best models for your tasks
• Potentially integrate additional data such as weather forecasts, GDP, ... into your forecasts to improve them