Modern Forecasting in Practice

4.7 (26)

·

3 Weeks

·

Cohort-based Course

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

With experience from

Amazon
Meta
Amazon Web Services
Zalando
SAP

Course overview

What you will learn

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.

Who is this course for

01

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

02

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

03

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

What you’ll get out of this course

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

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

Apply appropriate forecasting techniques to maximize the effectiveness of your solution

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

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

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


This course includes

8 interactive live sessions

Lifetime access to course materials

19 in-depth lessons

Direct access to instructor

Projects to apply learnings

Guided feedback & reflection

Private community of peers

Course certificate upon completion

Maven Satisfaction Guarantee

This course is backed by Maven’s guarantee. You can receive a full refund within 14 days after the course ends, provided you meet the completion criteria in our refund policy.

Course syllabus

Week 1

Mar 3—Mar 9

    Mar

    3

    Maven | Modern Forecasting in Practice TSF | Session 1

    Mon 3/35:00 PM—7:00 PM (UTC)

    Mar

    5

    Optional: Maven | Modern Forecasting in Practice TSF | Office Hours #1

    Wed 3/57:15 PM—7:45 PM (UTC)
    Optional

    Mar

    5

    Maven | Modern Forecasting in Practice TSF | Session 2

    Wed 3/55:00 PM—7:00 PM (UTC)

    Supplementary Materials

    3 items

    Materials from Lectures including homework

    2 items

    Recordings

    3 items

Week 2

Mar 10—Mar 16

    Mar

    12

    Optional: Maven | Modern Forecasting in Practice TSF | Office Hours #2

    Wed 3/127:15 PM—7:45 PM (UTC)
    Optional

    Mar

    12

    Maven | Modern Forecasting in Practice TSF | Session 4

    Wed 3/125:00 PM—7:00 PM (UTC)

    Mar

    10

    Maven | Modern Forecasting in Practice TSF | Session 3

    Mon 3/105:00 PM—7:00 PM (UTC)

    Mar

    13

    Optional: Deep Dive with Nixtla

    Thu 3/136:00 PM—7:00 PM (UTC)
    Optional

    Supplementary Materials (Week 2)

    3 items

    Recordings

    4 items

    Materials from Lectures including homework

    4 items

Week 3

Mar 17

    Mar

    24

    Maven | Modern Forecasting in Practice TSF | Session 5

    Mon 3/245:00 PM—7:00 PM (UTC)

4.7 (26 ratings)

What students are saying

What thought leaders are saying

        Jan and Tim are two of the best forecasters I know, especially when dealing with big data forecasting problems. Both have made important contributions to developing new machine-learning tools designed for forecasting and have years of relevant practical experience.
Rob Hyndman

Rob Hyndman

Professor of Statistics @ Monash University, Australia
        Over the past 10 years, Tim and Jan have tackled some of the hardest forecasting problems at Amazon and beyond. Their solutions advanced state-of-the-art and considered many aspects: from business questions during inception to method development and productization details during roll-out. Their rich and practical experience will benefit students.
Ralf Herbrich

Ralf Herbrich

Professor of Computer Science @ Hasso-Plattner Institute, Berlin
        I forecast that this course by Tim Januschowski and Jan Gasthaus on time series forecasting will be great.
Alex Smola

Alex Smola

Distinguished Scientist / VP @ Amazon Web Services

Meet your instructors

Tim Januschowski

Tim Januschowski

Director of Engineering, Databricks

is Director of Engineering and Site Lead Berlin for Databricks. Before joining Databricks, he was the Director of Pricing Platform at Zalando SE, where he leads the organization responsible for setting prices for the Zalando wholesale business. This involves forecasting of demand heavily. Prior to Zalando, Tim led the time series science organization for Amazon Web Services’ AI division. His teams built multiple AI services for AWS such as SageMaker, Forecast, Lookout for Metrics, and DevOps Guru, top-tier scientific publications, patents, and open source. Tim is a director at the International Institute of Forecasters, serves as a reviewer for the major ML venues, lectures at TU Munich, and advises start-ups such as WhyLabs.

Jan Gasthaus

Jan Gasthaus

Software Engineer, Meta

is a software engineer at Meta. Before that, he was principal machine learning scientist at Amazon, where he worked on some of the largest time series prediction problems on the planet. As part of AWS AI Labs, he helped create the technology behind AWS services such as Sagemaker, Amazon Forecast, and Amazon DevOps Guru, and co-created the open-source deep learning forecasting library GluonTS. Before building services for AWS, he worked on a wide range of forecasting and time series analysis problems across Amazon’s businesses, including the massive-scale retail demand forecasting problem, AWS capacity planning, workforce planning, price forecasting, and anomaly detection for cloud resources. Jan holds a Ph.D. in Machine Learning from UCL, has co-authored over 30 scientific articles on time series modeling, served as area chair and reviewer for NeurIPS and other major ML conferences, and has given numerous keynotes, lectures, and tutorials on forecasting.

A pattern of wavy dots

Join an upcoming cohort

Modern Forecasting in Practice

Cohort Q1 2025

$775

Dates

Mar 3—17, 2025

Payment Deadline

Apr 1, 2025
Get reimbursed

Course schedule for March 2024

4-6 hours per week

  • Session Timing & Details

    12:00pm - 2:00pm EST

    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.

  • Additional activities

    1 hour per week

    We'll have office hours, a deep dive with Max Mergenthaler, Nixtla and guest speaker Sercan Arik (Google). Past guest speakers:

    • Slawek Smyl (M4 winner; Walmart, Meta, Uber, Microsoft),
    • Sean Taylor (Prophet; Lyft, Meta)
    • Boris Oreshkin (NBEATS, Amazon)

What former participants are saying

        The opportunity to learn from Tim and Jan was great. I feel like it propelled me ahead a couple of years and it will help me avoid a lot of common pitfalls. I also think it helped me break out of some bad habits, like only looking at metrics and not paying a lot of attention to plots of (good and bad) forecasts. 
Boyd Biersteker

Boyd Biersteker

Data Scientist, RaboResearch
        I'm truly grateful for the course's practical approach in tackling challenging forecasting problems. The shared code and reading materials have been incredibly valuable. The course was enjoyable, and I'm already incorporating the gained knowledge into my daily work. Thank you for the enriching experience!
Eva Giannatou

Eva Giannatou

Senior Data Scientist, Navan
        The course had the perfect mix of theory and practice, preparing data scientists like me to tackle forecasting problems of any scale. Tim and Jan have prepared an excellent syllabus taking students from simple baselines and statistical time series models all the way to state of the art deep-learning methods.
Leonidas Tsaprounis

Leonidas Tsaprounis

Senior Data Scientist, Haelon

Learning is better with cohorts

Learning is better with cohorts

Active hands-on learning

This course builds on live workshops and hands-on projects

Interactive

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

Learn with a cohort of peers

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

Frequently Asked Questions

A pattern of wavy dots

Join an upcoming cohort

Modern Forecasting in Practice

Cohort Q1 2025

$775

Dates

Mar 3—17, 2025

Payment Deadline

Apr 1, 2025
Get reimbursed

$775

4.7 (26)

·

3 Weeks

Syllabus in more detail

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

Understand which & how business processes can be optimized by incorporating (probabilistic) predictions of future outcomes


Differentiate strategic from operational forecasting problems with examples from Zalando  and Amazon


Measure & compare the accuracy of different forecasts

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

Understand the underlying business problem & challenges of the resulting data-constrained forecasting problem


Identify the effects & structural components that make up the data, such as trend(s), seasonality, exexogenous shocks, noise


Identify the appropriate method and tools

Session 3: Forecasting solutions with a large 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 & tune global ML methods such as Gradient Boosted Trees and Neural Networks like DeepAR

Session 4: Forecasting solutions with dependency structures

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 & improve forecasting in counterfactual situations


Session 5: What best practices help you avoid common pitfalls in production? 

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


Stay in the loop

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

What people are saying

        Testimonials are a great way to add social proof to your landing page so that students feel more comfortable choosing your course.
Author name

Author name

Author title
        You can add as many as you’d like! We typically recommend landing somewhere in the 2-6 range and adding photos so that they feel more “real”.
Author name

Author name

Author title