Preface

Welcome to Time Series Analysis for Data Scientists! This online textbook grew out of the course notes for a time series analysis course at the University of Maryland, College Park given in the fall semester of 2025. This book aims to be largely self-contained with references to more advanced texts when necessary.

Background¶

I was motivated to write this book due to a frustration over not being able to find a text at a level I considered appropriate for data science students. On the one hand, there are multiple outstanding statistics texts covering the topic in great depth. However, the statistical and mathematical rigor of these texts can often cause students to become lost in the details without a solid understanding of what is being done and why. On the other hand, texts aimed at business students often go into too little depth for the purposes of data science and focus narrowly on econometrics applications at the expense of broader use cases relevant to data scientists.

This text aims to chart a middle course. Statistical and mathematical derivations are provided with the aim of equipping students with an understanding of the motivations, mechanisms, and limitations of various methods. However, we stop short of the level of rigor required for academic statisticians looking to derive novel methodologies. Where necessary, this book contains references to classics such as Shumway & Stoffer (2025) or Brockwell & Davis (1991) for more advanced students looking for additional depth. This book also contains topics not commonly covered in many time series classes that can be valuable for practicing data scientists such as SIR models from epidemiology and deep learning methodologies.

Most chapters contain one or more problems. Some are conceptual questions designed to help you put what you’ve learned into a broader context. Others are mathematical or statistical, designed to reinforce the underlying concepts. In the latter case, answers are generally provided in a hidden dropdown. I strongly encourage you to attempt the problems before looking at the solutions in order to gain the maximum benefit from this book.

Mathematical Level¶

This book is aimed at students who have a background in a quantitative discipline such as pure data science, physics, electrical engineering, or econometrics. The derivations generally assume at least two semesters of calculus and one semester of linear algebra as well as familiarity with the basics of probability and statistics. More advanced courses such as multi-variate calculus, Fourier analysis, and mathematical statistics are useful but not necessary.

Chapter 2 presents a refresher on several mathematical and statistical topics necessary for later chapters. Similarly, Chapter 5 covers an introduction to Fourier analysis. For many students it may be sufficient to skim the material to ensure familiarity. If you are not familiar with the topics the intention is that these chapters (with the accompanying worked problems) will serve as a sufficient introduction to master ensuing time series specific applications.

License¶

This book is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike license. In short, this means you are free to use this book, either in part or its entirety, for any non-commercial purposes provided you provide clear attribution. This work should be referenced as

Time Series Analysis for Data Scientists by Charles Forgy, PhD.

I would ask that any instructor or lecturer using substantial portions of the book please send me an email at ccforgy@umd.edu to help understand the contexts in which the book finds use and how it can be improved.

Contributing¶

This book is maintained at https://github.com/charlescforgy/time-series-book.git. If you find any errors or simply areas you’d like to see expanded upon I would be grateful if you open a pull request. If you are not comfortable with GitHub, you can also email me directly at ccforgy@umd.edu.

Anyone interested in getting more involved in this open source project to bring rigorous but applicable time series analysis tools to the public is invited to contribute. There are three areas in particular for which we are looking for contributors:

1. Copy editing for grammar and syntax

2. Translation into languages other than English

3. Graphic and website design to improve layout and readability

If you have other ideas for ways to improve the book feel free to drop me a line.

Best of luck in your time series journey!