Modern Aspects of Big Time Series Forecasting

Tutorial for IJCAI 2021

Time: 10:00-14:00, Thursday, August 19th, 2021
Location: Auditorium Red, Montreal-themed Virtual Reality, "Montreal"


Time series forecasting is a key ingredient in the automation and optimization of business processes: in retail, deciding which products to order and where to store them depends on the forecasts of future demand in different regions; in cloud computing, the estimated future usage of services and infrastructure components guides capacity planning; and workforce scheduling in warehouses and factories requires forecasts of the future workload. Recent years have witnessed a paradigm shift in forecasting techniques and applications, from computer-assisted model- and assumption-based to data-driven and fully-automated. This shift can be attributed to the availability of large, rich, and diverse time series data sources and result in a set of challenges that need to be addressed, such as the following: How can we build statistical models to efficiently and effectively learn to forecast from large and diverse data sources? How can we leverage the statistical power of “similar” time series to improve forecasts in the case of limited observations? What are the implications for building forecasting systems that can handle large data volumes?

The objective of this tutorial is to provide a concise and intuitive overview of the most important methods and tools available for solving large-scale forecasting problems. We review the state of the art in classical modeling of time series and modern methods with a particular focus on deep learning for forecasting. Furthermore, we discuss practical aspects of forecasting, evaluation and provide example problems. Our focus is on providing an intuitive overview of the methods and practical issues which we will illustrate via case studies. As an add-one we provide interactive material for self-study via Jupyter notebooks.


Video Recording



  • Introduction to time series forecasting and classical approaches (45 mins)
    • Modeling choices
    • Time series models: examples for local, univariate models
    • Linear models, exponential smoothing, and others
  • Deep learning forecasting: Part I (75 mins)
    • From linear regression to feed-forward networks
    • Basic model structure
    • Basic model blocks: RNN, CNN, Transformers
  • Break and Q&A (15 mins)

  • Deep learning forecasting: Part II (75 mins)
    • Generating probabilistic forecasts from NNs
    • Deep probabilistic models
  • Forecasting in practice (30 mins)
    • Getting started with forecasting

Presenters’ Bio

  Jan Gasthaus is a Senior Machine Learning Scientist in the Amazon AI Labs, working mainly on time series forecasting and large-scale probabilistic machine learning. He is passionate about developing novel machine learning solutions for addressing challenging business problems with scalable machine learning systems, all the way from scientific ideation to productization. Prior to joining Amazon, Jan obtained a BS in Cognitive Science from the University of Osnabrueck, an MS in Intelligent Systems from UCL, and a PhD from the Gatsby Unit, UCL, focusing on Nonparametric Bayesian methods for sequence data.


  Tim Januschowski is a Machine Learning Science Manager in Amazon Web Services’ AI Labs where he leads science teams across the world working on machine learning science for time series. Algorithms by the group run in production both internally at Amazon and are available to AWS customers through AWS services such as Amazon Forecast, Amazon Lookout for Metrics, Amazon SageMaker and Amazon DevOpsGuru. At the same time, he publishes at the major ML outlets in time series applications such as anomaly detection and forecasting. He studied Mathematics at TU Berlin, IMPA, Rio de Janeiro, and Zuse-Institute Berlin and holds a PhD from University College Cork. He serves as a director at the International Institute of Forecasters.


  Yuyang (Bernie) Wang is a Principal Machine Learning Scientist in Amazon AI Labs, working mainly on large-scale probabilistic machine learning with its application in Forecasting. He received his PhD in Computer Science from Tufts University, MA, US and he holds an MS from the Department of Computer Science at Tsinghua University, Beijing, China. His research interests span statistical machine learning, numerical linear algebra, and random matrix theory. In forecasting, Yuyang has worked on all aspects ranging from practical applications to theoretical foundations.


Some recent tutorials by Christos Faloutsos and Co. on big time series mining:

Getting Started with GluonTS