Modern Time Series Forecasting with Python: Explore industry-ready time series forecasting using modern machine learning and deep learning

It covers almost every important topics and the code provided by the author is really helpful. Easy to follow and great inspiration.

The book is about 500+ pages long. But the text is in smaller font than other books so basically the book can be 50% bigger if it was in a slightly larger font. In addition, Manu has left out all the code he has written from the book (but you’ll find it in the GitHub for the book and it’s free!). If he had published all the code that he uses and references in the book, the book can easily be 1500 pages long. That why I call it massive!Why do I like this book so much?This book is very detailed, very thorough when it comes to explaining the huge number of techniques that one uses in forecasting from ARIMA to exponential smoothing to ML to deep learning and now transformers. Manu does not leave anything behind.In addition, Manu explains how to overcome hurdles with each technique by explaining their pitfalls and takes you along on a joyous ride from beginning to end.I have gone through the most exciting chapters for me which begin from chapter 8 through 17. These address some of the most common techniques we use in forecasting in the real world. The book provides example notebooks for each technique from Linear regression to temporal fusion transformers (TFT). The interesting thing is Manu has coded almost all of the notebooks himself.So I thought I will give everyone a heads-up about this fantastic book while I spend a month or so finishing it. Until then, if you can afford it and have the need to solve forecasting problems in your work, make sure you buy this book or get your company to pay for it. It’ll save you tons of time and money! Highly recommended!#ml #deeplearning #forecasting #pytorch

If you want to develop an understanding and expertise in time series, this post is for you!The book provides a 360 degrees view of time series modeling. This book has everything from essential components of time series modeling to deep learning models.Key Highlights:1. A detailed explanation of different kinds of forecasting, such as multivariate forecasting, and in-sample/out-sample, is crucial to understanding picking the right kind of modeling approach.2. Detailed explanation on how to perform data wrangling to prepare your data for a successful time series modeling model, like treating missing values.3. Proper explanation of Brain functioning and Linear Algebra before diving deep into the Neural Network Model. The author has provided accurate intuition to set you up for success, even if you are reading about Deep Learning for the first time.Some new concepts that you might learn:1. Temporal embeddings and time delay embeddings and the benefits of creating global models over the local model2. What is data leakage and ways to protect it3. Multi-task learning, like Global Forecasting Models4. Novel ways to validate your forecast beyond traditional metricsThe book is perfect for:1. Data Science Practitioners looking to upgrade their knowledge of time series modeling2. Aspiring Data Scientists looking to initiate their knowledge on the subject3. You are an expert in ML aspects of Time Series but want to gain expertise using Deep Learning for forecasting and vice versa.As a bonus, this book has python codes and practitioner tips in most chapters. The book reflects Manu Joseph's expertise and grasp of theoretical concepts.

Of several recent books on time series and forecasting in Python, this is arguably the best. The introductory chapters deal with several key issues in modern time series analysis. It is understood that real world time series exhibit properties such as serial correlation, cyclicality and nonstationarity, which are rarely addressed in older textbooks. The sections on classical forecasting methods include not only linear regression but exponential smoothing, ARIMA and spectral analysis. The sections on artificial intelligence include newer techniques such as recurrent neural nets. The book deals with forecasting over multiple intervals, a topic that is more complex in AI applications than might at first sight be apparent. Although some very advanced approaches are not dealt with in detail, this book provides an excellent introduction. A nice feature is the inclusion of sample code.

I found this book to be a priceless resource as a student who has three years of expertise in data science and machine learning and is currently doing a master's in AI. I've been practically applying the concepts of this book to realtime industrial capstone projects. The book offers a rich explanation of time series forecasting that is written in an easy to digest manner. It offers a thorough review of all the crucial facets of time series modeling and forecasting for people who are just beginning their careers in machine learning. The book also offers machine learning experts a road map for using the most recent deep-learning methods to improve their time series models. The PyTorch Forecasting library is covered throughout the book in a self-contained and simple-to-follow manner, especially for people who are new to PyTorch, which makes it a very helpful resource. I've been testing these methods in my regular work and getting fantastic outcomes.The book's layout is carefully thought out, with four main sections that progress from a fundamental introduction to time series to the usage of deep learning architectures. The material of each chapter is made simple yet insightful to follow through the use of mathematical illustrations, graphs, flow diagrams, tables, photos, and code snippets.I would like to extend my heartfelt congratulations to Manu for sharing his expertise and insights on a broad range of concepts that are relevant to solving the challenges faced by industry professionals today.