Learn traditional and cutting-edge machine learning (ML) and deep learning techniques and best practices for time series forecasting, including global forecasting models, conformal prediction, and transformer architectures

Key Features

• Apply ML and global models to improve forecasting accuracy through practical examples
• Enhance your time series toolkit by using deep learning models, including RNNs, transformers, and N-BEATS
• Learn probabilistic forecasting with conformal prediction, Monte Carlo dropout, and quantile regressions

Book Description

Predicting the future, whether it's market trends, energy demand, or website traffic, has never been more crucial. This practical, hands-on guide empowers you to build and deploy powerful time series forecasting models. Whether you’re working with traditional statistical methods or cutting-edge deep learning architectures, this book provides structured learning and best practices for both.

Starting with the basics, this data science book introduces fundamental time series concepts, such as ARIMA and exponential smoothing, before gradually progressing to advanced topics, such as machine learning for time series, deep neural networks, and transformers. As part of your fundamentals training, you’ll learn preprocessing, feature engineering, and model evaluation. As you progress, you’ll also explore global forecasting models, ensemble methods, and probabilistic forecasting techniques.

This new edition goes deeper into transformer architectures and probabilistic forecasting, including new content on the latest time series models, conformal prediction, and hierarchical forecasting. Whether you seek advanced deep learning insights or specialized architecture implementations, this edition provides practical strategies and new content to elevate your forecasting skills.

What you will learn

• Build machine learning models for regression-based time series forecasting
• Apply powerful feature engineering techniques to enhance prediction accuracy
• Tackle common challenges like non-stationarity and seasonality
• Combine multiple forecasts using ensembling and stacking for superior results
• Explore cutting-edge advancements in probabilistic forecasting and handle intermittent or sparse time series
• Evaluate and validate your forecasts using best practices and statistical metrics

Who this book is for

This book is ideal for data scientists, financial analysts, quantitative analysts, machine learning engineers, and researchers who need to model time-dependent data across industries, such as finance, energy, meteorology, risk analysis, and retail. Whether you are a professional looking to apply cutting-edge models to real-world problems or a student aiming to build a strong foundation in time series analysis and forecasting, this book will provide the tools and techniques you need. Familiarity with Python and basic machine learning concepts is recommended.

Table of Contents

1. Introducing Time Series
2. Acquiring and Processing Time Series Data
3. Analyzing and Visualizing Time Series Data
4. Setting a Strong Baseline Forecast
5. Time Series Forecasting as Regression
6. Feature Engineering for Time Series Forecasting
7. Target Transformations for Time Series Forecasting
8. Forecasting Time Series with Machine Learning Models
9. Ensembling and Stacking
10. Global Forecasting Models
11. Introduction to Deep Learning
12. Building Blocks of Deep Learning for Time Series
13. Common Modeling Patterns for Time Series

(N.B. Please use the Read Sample option to see further chapters)

Review

“Great to see Jeffrey Tackes joining forces with Manu Joseph to finally bring the book about forecasting by the people who are actually building large scale forecasting systems for real companies.”

Valeriy Manokhin, Head of Data Science, Author of Practical Guide to Applied Conformal Prediction

“Suffices to say, the first edition of this book is an absolute must-have—and now a second edition is on the way! The only thing I thought was missing from the first edition was coverage of probabilistic forecasting; well, this has now been remedied in the form of Chapter 17, which is over 60 pages long. [...] The authors discuss the latest progress in addressing the inherent exchangeability problem in time series data, and when I say latest progress, they even refer to research papers that are barely a few months old.”

Carl McBride Ellis, PhD, Predictive Analytics, Author of The Orange Book of Machine Learning

About the Author

Manu Joseph is a self-made data scientist with more than a decade of experience working with many Fortune 500 companies enabling digital and AI transformations, specifically in machine learning-based demand forecasting. He is considered an expert, thought leader, and strong voice in the world of time series forecasting. Currently, Manu leads applied research at Thoucentric, where he advances research by bringing cutting-edge AI technologies to the industry. He is also an active open-source contributor and developed an open-source library—PyTorch Tabular—which makes deep learning for tabular data easy and accessible. Originally from Thiruvananthapuram, India, Manu currently resides in Bengaluru, India, with his wife and son

Jeff Tackes is a seasoned data scientist specializing in demand forecasting with over a decade of industry experience. Currently he is at Kraft Heinz, where he leads the research team in charge of demand forecasting. He has pioneered the development of best-in-class forecasting systems utilized by leading Fortune 500 companies. Jeff's approach combines a robust data-driven methodology with innovative strategies, enhancing forecasting models and business outcomes significantly. Leading cross-functional teams, Jeff has designed and implemented demand forecasting systems that have markedly improved forecast accuracy, inventory optimization, and customer satisfaction. His proficiency in statistical modeling, machine learning, and advanced analytics has led to the implementation of forecasting methodologies that consistently surpass industry norms. Jeff's strategic foresight and his capability to align forecasting initiatives with overarching business objectives have established him as a trusted advisor to senior executives and a prominent expert in the data science domain. Additionally, Jeff actively contributes to the open-source community, notably to PyTimeTK, where he develops tools that enhance time series analysis capabilities. He currently resides in Chicago, IL with his wife and son.