Time Series Forecasting in Python: End-to-End Practice
Build, benchmark, and evaluate a full forecasting pipeline using ML, deep learning, and GenAI
Hands-on, end-to-end workshop | Real dataset (M5) | ML, Deep Learning & GenAI benchmarking
This is a 4-hour, hands-on live workshop designed for practitioners who want to move beyond isolated models and build a complete, production-ready forecasting workflow in Python.
You will implement an end-to-end pipeline using the Nixtla ecosystem and a curated subset of the M5 dataset - starting from strong statistical baselines and progressing to machine learning, deep learning, and modern foundation (GenAI-style) models.
Beyond modeling, the workshop focuses on how forecasting actually works in practice: how to frame problems correctly, design robust backtesting strategies, evaluate models beyond accuracy, and decide what is worth deploying in real-world systems.
You will also learn how to benchmark different model families in a single workflow, understand when each approach works best, and compare performance across speed, accuracy, and practicality.
By the end of the workshop, you will have built and compared multiple forecasting approaches and developed a clear, practical framework for selecting models and making forecasting decisions under uncertainty.
Who this is for
- Data scientists working with time series data
- ML engineers building predictive systems
- Forecasting practitioners exploring modern and GenAI-based approaches
- Supply chain and demand planning analysts
- Analytics professionals responsible for forecasting decisions
What You Will Learn
- How to build a complete, end-to-end forecasting pipeline in Python (from baselines to modern models)
- How to benchmark classical, ML, deep learning, and GenAI-style models in a single workflow
- How to design robust backtesting strategies and evaluate models beyond accuracy
- How to generate and interpret prediction intervals to reason about uncertainty and risk
- How to select the right forecasting approach based on real-world data patterns and constraints
- Hands-on experience with modern forecasting tools, including the Nixtla ecosystem
What You’ll get
- Free bestselling eBook (Modern Time Series Forecasting with Python, 2nd Edition)
- Complete, ready-to-use Python notebooks (end-to-end pipeline)
- Template code for ML, deep learning, and foundation model benchmarking
- Forecasting cheat sheet (models, metrics, decision framework)
- Step-by-step guide for prediction intervals and evaluation
- Reusable feature engineering and backtesting templates
- Curated resources for further learning
- Event recording and certificate
Pre-requisites
- Intermediate proficiency in Python
- Familiarity with pandas and basic machine learning concepts
- Access to Google Colab or a local Python environment
- (Optional) Basic exposure to time series forecasting
Build, benchmark, and evaluate a full forecasting pipeline using ML, deep learning, and GenAI
Hands-on, end-to-end workshop | Real dataset (M5) | ML, Deep Learning & GenAI benchmarking
This is a 4-hour, hands-on live workshop designed for practitioners who want to move beyond isolated models and build a complete, production-ready forecasting workflow in Python.
You will implement an end-to-end pipeline using the Nixtla ecosystem and a curated subset of the M5 dataset - starting from strong statistical baselines and progressing to machine learning, deep learning, and modern foundation (GenAI-style) models.
Beyond modeling, the workshop focuses on how forecasting actually works in practice: how to frame problems correctly, design robust backtesting strategies, evaluate models beyond accuracy, and decide what is worth deploying in real-world systems.
You will also learn how to benchmark different model families in a single workflow, understand when each approach works best, and compare performance across speed, accuracy, and practicality.
By the end of the workshop, you will have built and compared multiple forecasting approaches and developed a clear, practical framework for selecting models and making forecasting decisions under uncertainty.
Who this is for
- Data scientists working with time series data
- ML engineers building predictive systems
- Forecasting practitioners exploring modern and GenAI-based approaches
- Supply chain and demand planning analysts
- Analytics professionals responsible for forecasting decisions
What You Will Learn
- How to build a complete, end-to-end forecasting pipeline in Python (from baselines to modern models)
- How to benchmark classical, ML, deep learning, and GenAI-style models in a single workflow
- How to design robust backtesting strategies and evaluate models beyond accuracy
- How to generate and interpret prediction intervals to reason about uncertainty and risk
- How to select the right forecasting approach based on real-world data patterns and constraints
- Hands-on experience with modern forecasting tools, including the Nixtla ecosystem
What You’ll get
- Free bestselling eBook (Modern Time Series Forecasting with Python, 2nd Edition)
- Complete, ready-to-use Python notebooks (end-to-end pipeline)
- Template code for ML, deep learning, and foundation model benchmarking
- Forecasting cheat sheet (models, metrics, decision framework)
- Step-by-step guide for prediction intervals and evaluation
- Reusable feature engineering and backtesting templates
- Curated resources for further learning
- Event recording and certificate
Pre-requisites
- Intermediate proficiency in Python
- Familiarity with pandas and basic machine learning concepts
- Access to Google Colab or a local Python environment
- (Optional) Basic exposure to time series forecasting
Lineup
Jeffrey Tackes
Manu Joseph
Good to know
Highlights
- 4 hours
- Online
Refund Policy
Location
Online event
Agenda
Welcome and Forecasting Workflow Overview (5 mins)
We begin with a quick setup and introduction to the end-to-end time series forecasting workflow you will build during the workshop. This session frames how modern forecasting has evolved—from classical statistical methods to machine learning, deep learning, and emerging GenAI-style foundation models—and how these approaches fit together in real-world systems.
Data Schema, Problem Framing, and Evaluation Strategy (20 mins)
This session focuses on structuring time series data and defining the forecasting problem correctly. You will learn how to set forecast horizons, design robust backtesting strategies, and evaluate models using both point accuracy and uncertainty metrics. We also introduce pooled evaluation techniques and discuss why they matter for scalable, production-grade forecasting systems.
Fast EDA and Time Series Data Diagnostics (15 mins)
Before modeling, understanding your data is critical. You will perform fast exploratory data analysis (EDA) to identify key patterns such as seasonality, sparsity, zeros, and intermittency. This session emphasizes how these patterns influence model selection across classical methods, ML models, and modern GenAI-inspired forecasting approaches.