11 chapters. 762 pages. 264 references. 119 figures. From foundations to foundation models — in one book.
This is the forecasting reference that didn’t exist until now.
Most forecasting books fall into one of two traps: either they rehash ARIMA and exponential smoothing as if deep learning never happened, or they throw transformer architectures at you with no evaluation discipline and no explanation of when classical methods still win. This book does neither.
Mastering Modern Time Series Forecasting is an end-to-end, practitioner-first guide to building forecasting systems that hold up in production. It covers the full stack — classical statistical models, feature-engineered ML, deep learning, transformers, and modern foundation time series models — with the evaluation and validation rigour that most books skip entirely.
Written by Valery Manokhin, PhD, MBA, CQF. PhD in Machine Learning from Royal Holloway, University of London, supervised by Professor Vladimir Vovk (the creator of Conformal Prediction). MSc in Computational Statistics and Machine Learning from UCL. Years of building and auditing production forecasting systems — and seeing exactly how projects fail when evaluation is wrong, assumptions are ignored, or shiny models replace engineering discipline.
