New time‑series forecasting book
A newly published book titled 'Mathematics of Time Series Forecasting' lays out Python implementations of statistical, machine‑learning and deep‑learning methods for forecasting. The book is positioned as a practical guide you can follow to implement and compare forecasting approaches relevant to trading and product metrics. (x.com)
A new book published on March 23, 2026, packages time-series forecasting into one Python guide that runs from classical statistics to deep learning. (amazon.com) Time-series forecasting is the practice of using dated observations — sales by week, prices by minute, traffic by hour — to estimate what comes next. Forecasting texts often split those patterns into trend, seasonality, and residual noise before modeling them. (otexts.com) The book, *Mathematics of Time Series Forecasting*, is listed under ISBN 9789349887664 and credited to Dr. Sulekha AloorRavi. Amazon and Apple Books both show it as a new release from late March 2026. (amazon.com) (books.apple.com) Retail listings describe a 278-page manual built around Python implementations and step-by-step derivations. The methods named in those listings include AutoRegressive Integrated Moving Average, Seasonal AutoRegressive Integrated Moving Average, Exponential Smoothing, Vector AutoRegression, gradient boosting, neural networks, and Long Short-Term Memory models. (amazon.in) (wowebook.org) Those model families solve different parts of the same problem. AutoRegressive Integrated Moving Average and Exponential Smoothing are standard baselines for recurring patterns, while gradient boosting is widely used for tabular prediction and TensorFlow documents Long Short-Term Memory as a sequence model for ordered data such as time series. (otexts.com) (scikit-learn.org) (tensorflow.org) The practical pitch is comparison, not allegiance to one method. Scikit-learn’s documentation shows gradient boosting can be used for regression and for quantile-based prediction intervals, while TensorFlow’s time-series tutorial walks through single-step and multi-step forecasting with recurrent and convolutional neural networks. (scikit-learn.org 1) (scikit-learn.org 2) (tensorflow.org) That framing matches a broader shift in forecasting education toward Python-based workflows. Rob Hyndman and George Athanasopoulos’s *Forecasting: Principles and Practice* now has a Python edition online, reflecting demand for the same concepts in the software stack many machine-learning teams already use. (otexts.com) The audience appears to be practitioners who need forecasts tied to business or market data rather than a pure theory text. Multiple listings say the book covers data cleaning, decomposition, feature engineering, validation, and implementations relevant to trading signals and product metrics. (amazon.ca) (amazon.sg) The release lands into a crowded shelf of forecasting books, but the distinguishing claim is the mix of mathematical derivation and runnable Python in one volume. For readers choosing between classical models and newer neural approaches, the book is being sold as a side-by-side manual rather than a manifesto. (amazon.com 1) (amazon.com 2)
