Cover Image for Time Series Bootcamp: Forecasting with ML, Deep Learning, Foundation Models
Cover Image for Time Series Bootcamp: Forecasting with ML, Deep Learning, Foundation Models
Avatar for Packt - ML & AI Engineering

Time Series Bootcamp: Forecasting with ML, Deep Learning, Foundation Models

Virtual
Get Tickets
Welcome! Please choose your desired ticket type:
Full workshop access
 
Participants will receive a full HD recording of both days, the workshop slide deck, a certificate of participation, and practical Python notebooks covering classical forecasting methods, machine learning, deep learning, and foundation models. They will also take away a reusable forecasting evaluation harness, model implementations, a common scoring and leaderboard system, and a practical framework for comparing and selecting models based on accuracy, stability, business costs, and forecasting requirements. Through hands-on exercises using real-world retail data, attendees will gain experience with hyperparameter tuning, exogenous variables, ensembling, uncertainty quantification, and hierarchical reconciliation, along with insights from three expert instructors and opportunities for live Q&A.
Available until  
Day-1 Workshop Access
 
This pass provides access to Day 1 only of the Time Series Forecasting Bootcamp. Build baselines, compare machine learning models, and explore price and promotion effects using MLForecast and Nixtla. Practise feature preparation, forecasting across multiple horizons, model tuning, and performance evaluation. Review leaderboard results to inform model selection. Day 2 access is not included with this ticket.
Available until  
Day-2 Workshop Access
 
This pass provides access to Day 2 only of the Time Series Forecasting Bootcamp. Train and compare neural forecasting models and foundation models. Explore tuning, price and promotion effects, uncertainty, and forecasts across product groups. Compare individual models with combined approaches, then use leaderboard results to guide model selection. Day 1 access is not included with this ticket.
Available until  
About Event

​Overview

​Build, benchmark, and select time series forecasting models using real-world retail data in a two-day, hands-on bootcamp, with 4 to 5 hours of learning each day.

​Workshop description

​Work through a shared retail dataset featuring intermittent demand, promotions, long-tail products, and product hierarchies. Compare statistical, machine learning, deep learning, and foundation models using a common evaluation framework.

​Practical labs and a shared leaderboard help you assess accuracy, bias, stability, uncertainty, computational costs, and business impact.

​Who this workshop is for

  • ​Data scientists and ML engineers developing forecasting solutions.

  • ​Forecasting analysts, demand planners, and supply-chain professionals with Python experience.

  • ​Researchers and analytics practitioners evaluating modern forecasting methods.

​By the end, you will be able to build

  • ​Forecasting pipelines with relevant features, exogenous variables, and leakage prevention.

  • ​Time-aware backtesting and evaluation workflows.

  • ​Statistical, gradient-boosted, neural, and foundation-model forecasting experiments.

  • ​Prediction intervals and forecasts reconciled across product hierarchies.

  • ​Model comparisons that support selecting individual models or ensembles for business requirements.

​Recommended preparation:

  • ​Working knowledge of Python and data analysis.

  • ​Basic familiarity with machine learning and time series concepts.

  • ​Ability to run Python notebooks in Google Colab.

​Key takeaways

  • ​Reusable forecasting notebooks, a GitHub repository, and an evaluation harness.

  • ​A practical framework for balancing forecast quality, business costs, and computational requirements.

  • ​Experience with StatsForecast, MLForecast, NeuralForecast, TimeCopilot, Chronos-2, and TabPFN-TS.

  • ​Full HD recording, presentation slides, and a Certificate of Completion.

  • ​Direct Q&A with three forecasting practitioners.

​About the instructors

  • ​Graeme Davidson: Lead Data Scientist at Retail Express and co-author of Time Series with PyTorch. Brings practical experience connecting forecasting methods with retail decisions.

  • ​Lei Ma: Physicist-turned-data scientist and co-author of Time Series with PyTorch. Specializes in deep learning and forecasting across several industries.

  • ​Jeffrey Tackes: Founder of Forecast Academy and co-author of Modern Time Series Forecasting with Python, Second Edition. Focuses on practical backtesting, model evaluation, and demand planning.

Avatar for Packt - ML & AI Engineering