Cohort Q3 2024

Get hands-on experience with modern forecasting tools & learn from case studies of the toughest forecasting challenges in the industry

Tim Januschowski

Director of Engineering, Databricks

Jan Gasthaus

Software Engineer & Applied Scientist

Modern Forecasting in Practice

Amazon

Amazon Web Services

Zalando

Learn how to use machine learning techniques for predicting future outcomes in time series to optimize business processes. The course features practical lessons heavily that we derived from two decades of working on some of the world's hardest forecasting problems at Amazon retail, Zalando and for AWS and its customers. You'll pick up the necessary theory, get hands-on example and learn about the tricks of the forecasting trade.

What you will learn

Data Scientists who want to go beyond standard ML/AI problems and solve forecasting related business problems

Business analysts with familiarity in machine learning in industry settings who want to uplevel themselves in a top ML application domain

Economists and Applied Scientists who want to apply industry-proven modern forecasting techniques

Who is this course for

What you’ll get out of this course

<ul><li>Understand which and how business processes can be optimized by (probabilistic) forecasts</li><li>Differentiate strategic from operational forecasting problems with real-world examples from Zalando and Amazon</li><li>Measure and compare the accuracy of different forecasts</li></ul>

Identify the business problems that can benefit from modern time series forecasting techniques
 

<ul><li>obtain the right co-variates for your problems and learn how to process them</li><li>pick the appropriate model family (and software implementation) for your forecasting challenge</li><li>enable model iterations through tools such as state of the art hyper parameter optimization</li></ul>

Apply appropriate forecasting techniques to maximize the effectiveness of your solution
 

<ul><li>Visualize results in meaningful ways to make your efforts tangible to a non-technical audience</li><li>Understand the right quantitative evaluations to ensure that your results are convincing</li></ul>

Develop the skills to present results effectively to persuade a non-technical audience 
 

<ul><li>Understand which and how business processes can be optimized by incorporating (probabilistic) predictions of future outcomes</li><li>Differentiate strategic from operational forecasting problems with examples from Zalando and Amazon</li><li>Measure and compare the accuracy of different forecasts</li></ul>

Session 1: Should Your Business Problem Be Solved With Forecasting?

<ul><li>Case Study: Forecasting top-level energy demand and prices</li><li>Understand the underlying business problem and the challenges of the resulting data-constrained forecasting problem</li><li>Identify the effects and structural components that make up the data, such as trend(s), seasonality, exogenous shocks, and noise</li><li>Identify the appropriate method and tool, such</li></ul>

Session 2: Forecasting Solutions Using a Small Set of Time Series

 

<ul><li>Case Study: Retail demand forecasting</li><li>Build an intuition for the data via visualization of individual time series and aggregate summaries</li><li>Obtain co-variates/features and process them</li><li>Use and tune global ML-powered methods such as Gradient Boosted Trees and Neural Network-based methods like DeepAR</li></ul>

Session 3: Forecasting Solutions With a Large Set of Time Series

 

<ul><li>Case Study: Forecasting with causal inputs</li><li>Forecast demand subject to price changes for millions of products</li><li>Build what-if analysis using simple and advanced approaches</li><li>Evaluate &amp; improve forecasting in counterfactual situations</li></ul>

Session 4: Forecasting Solutions With Dependency Structures

 

<ul><li>Practical tactics for forecasting exemplified by labor planning</li><li>Productionize forecasting models including retraining schemes</li><li>Handle missing data and the associated perils</li><li>Research approaches to outliers/extreme events such as blizzards and pandemics</li></ul>

Session 5: What Best Practices Help You Avoid Common Pitfalls in Production?

Course syllabus

Professor of Statistics @ Monash University, Australia

Professor of Computer Science @ Hasso-Plattner Institute, Berlin

Distinguished Scientist / VP @ Amazon Web Services

What thought leaders are saying

Meet your instructors

We have 5 full sessions each 2h, starting on Monday 4 March, 12pm EST (6pm CET). Sessions will consist of an engaging mix of presentations, activities and notebooks.

Session Timing & Details

We'll have office hours, a deep dive with Max Mergenthaler, Nixtla and guest speaker Sercan Arik (Google). Past guest speakers:<ul><li>Slawek Smyl (M4 winner; Walmart, Meta, Uber, Microsoft),</li><li>Sean Taylor (Prophet; Lyft, Meta)</li><li>Boris Oreshkin (NBEATS, Amazon)</li></ul>

Additional activities

Course schedule for March 2024

What former participants are saying

This course builds on live workshops and hands-on projects

Active hands-on learning

You’ll be interacting with other learners through breakout rooms and peer exchange

Interactive 

Join a community of like-minded people who want to learn and grow alongside you

Learn with a cohort of peers

Learning is better with cohorts

Frequently Asked Questions

Sign up to be the first to know about course updates.

Stay in the loop

Syllabus in more detail

Understand which &amp; how business processes can be optimized by incorporating (probabilistic) predictions of future outcomes Differentiate strategic from operational forecasting problems with examples from Zalando &nbsp;and Amazon Measure &amp; compare the accuracy of different forecasts

Session 1: Should your business problem be solved with forecasting? 

Understand the underlying business problem &amp; challenges of the resulting data-constrained forecasting problem Identify the effects &amp; structural components that make up the data, such as trend(s), seasonality, exexogenous shocks, noise Identify the appropriate method and tools

Session 2: Forecasting solutions using a small set of time series 

Case Study: Retail demand forecasting Build an intuition for the data via visualization of individual time series and aggregate summaries Obtain co-variates/features and process them Use &amp; tune global ML methods such as Gradient Boosted Trees and Neural Networks like DeepAR

Session 3: Forecasting solutions with a large set of time series

Case Study: Forecasting with causal inputs Forecast demand subject to price changes for millions of products Build what-if analysis using simple and advanced approaches Evaluate &amp; improve forecasting in counterfactual situations

Session 4: Forecasting solutions with dependency structures

Practical tactics for forecasting exemplified by labor planning Productionize forecasting models including retraining schemes Handle missing data and the associated perils Research approaches to outliers/extreme events such as blizzards and pandemics

Course completed by

About this course

Get notified about the next cohort

Learn from live sessions