Machine Learning for Time Series Forecasting with Python

Forecasting with time series data is not just a matter of choosing an algorithm. The work also involves understanding the data, preparing it thoughtfully, judging how well a model performs, and getting a useful solution into operation. Francesca Lazzeri’s Machine Learning for Time Series Forecasting with Python follows that broader path, pairing forecasting methods with the practical steps around them.

From the data to a forecasting plan

The opening chapters introduce time series forecasting and show how to shape an end-to-end solution. Along the way, the book addresses business understanding and performance metrics, data ingestion and exploration, preprocessing, feature engineering, model selection, and evaluation. A demand-forecasting use case brings the workflow into focus.

Prepare time series data with care

Readers encounter the practical details that can influence a forecast: timestamps and periods, indexing, frequency conversion, missing values, and normalization. The data-preparation material also covers date-time features, lag and window features, and rolling or expanding statistics—useful building blocks when turning sequential observations into model-ready inputs.

Compare established methods and neural networks

The book introduces autoregression, moving averages, ARMA and ARIMA, alongside automated machine learning. It then turns to neural-network approaches for forecasting, including recurrent neural networks, long short-term memory (LSTM) networks, and gated recurrent units (GRUs), with discussion of univariate and multivariate models.

Carry the model toward deployment

Model deployment receives its own chapter, including Azure Machine Learning SDK for Python concepts and a worked deployment path for an ARIMA model. This makes the book’s scope broader than model theory alone: it considers how forecasting work proceeds from data and experiments toward a deployed service.

A practical reference for forecasting work

With its combination of data preparation, traditional forecasting methods, neural networks, evaluation, and deployment, this book can help data scientists, analysts, developers, and researchers build a more systematic view of machine-learning-based forecasting. It is particularly relevant to readers who want to understand how the stages of a forecasting project fit together.