Mastering Advanced Time Series Forecasting in Python
Probabilistic, Hierarchical, and Foundation Models
This book is the advanced sequel to Mastering Modern Time Series Forecasting.
It is written for practitioners who already know classical and machine-learning forecasting methods and now need to build robust, production-grade forecasting systems.
The focus is on probabilistic forecasting, hierarchical coherence, and modern foundation models for time series, with clear Python implementations and real-world constraints in mind.
What You Will Learn
This book covers advanced time series forecasting topics that matter in production:
- Probabilistic forecasting
- Prediction intervals and distributions
- Conformal prediction and calibration
- When uncertainty estimation adds value—and when it does not
- Hierarchical time series forecasting
- Forecast reconciliation and coherence
- Bottom-up, top-down, and hybrid approaches
- Applications to supply chains, portfolios, and planning systems
- Foundation models for time series
- Transformer-based models and large time-series models
- Strengths, limitations, and failure cases
- Evidence-based evaluation beyond hype
- Forecasting system design
- Forecastability assessment
- Metric selection and optimization
- Model selection under business constraints
- Monitoring, drift, governance, and MLOps
All examples use Python and emphasize reproducibility, robustness, and scalability.
How This Book Is Different
This is not a cookbook and not a Kaggle guide.
You will learn:
- Why specific forecasting methods work or fail
- Trade-offs between accuracy, stability, and interpretability
- Common anti-patterns in real forecasting projects
- How to communicate forecasts and uncertainty to stakeholders
The emphasis is on engineering forecasting systems, not just training models.
Who This Book Is For
- Data Scientists and ML Engineers working on time series forecasting
- Quantitative analysts in finance, energy, retail, and operations
- Senior ICs and technical leads responsible for forecasting decisions
- Readers familiar with ARIMA, ETS, and basic ML who want advanced methods
This is not an introductory time series book.
Leanpub Living Book
This is a Leanpub living book:
- New chapters added over time
- Continuous corrections and improvements
- Additional tools, templates, and case studies
All updates are included free for life after purchase.
About the Author
Valeriy Manokhin, PhD, MBA, CQF is an independent forecasting researcher and practitioner.
He has built large-scale forecasting systems for global companies, outperforming major consultancies and specialized AI vendors. His work has delivered measurable business impact across industry and academia, with readers and students in over 100 countries.
Optional: Live Training
Readers can complement the book with live instruction:
Modern Forecasting Mastery (Maven)
https://maven.com/valeriy-manokhin/modern-forecasting-mastery
What You Get
- Immediate access to all current chapters
- Free lifetime updates as the book evolves
