Review of Stochastic Modeling: A Thorough Guide to Evaluate, Pre-Process, Model and Compare Time Series with MATLAB Software by Hossein Bonakdari and Mohammad Zeynoddin

Data mining and modeling are two subjects with vast applications in all fields of science and industry. Regarding their importance in developing knowledge, they are considered an interdisciplinary science; therefore, most researchers and business owners use these methods to conduct analysis and enhance their work. Computational modeling and time series analysis methods have changed since Box and Jenkins (1976) introduced the fundamentals of statistical modeling. With the growth of the power of computational units, various artificial intelligence (AI) and developments in statistical methods were presented in 1976 to 2024. Many books with similar themes have been published, tutoring these methods and providing conceptual content. Despite different advantages and disadvantages, they mostly have one feature in common: paying too much attention to theoretical concepts. Providing details in its place is very useful; still, such information can sometimes be misleading for readers. Even the complexity of these concepts could prevent beginners in this field from continuing. Thus, a book that explains the concepts of data analysis and presents stochastic methods, data pre- and postprocessing methods, and even combining different techniques in simple terms was utterly needed. This book could also provide the easiest programming method for developing these methods. Bonakdari and Zeynoddin provide a simplified and comprehensive description of these theories and present the pathway to data modeling goals step by step with an easy programming language.

This book is a complete manifestation of research data mining and real-world phenomena time series modeling that started with Bonakdari et al. (2019). Thereafter, the investigation focused on diverse aspects of this field. This research answered serious questions, and the gaps felt in the domain of time series modeling were filled. The ongoing inquiry of stochastic methods or AI methods was responded to by Ebtehaj et al. (2020), and the different aspects of data preprocessing were investigated by Zeynoddin et al. (2018). This path was enriched as they published many other articles. These subjects are represented in the book and provide a clear path for readers.

The problem with real-life data sets is that they have sophisticated structures. Many complexities are associated with the data distribution, long-term or short-term underlying patterns, and even redundant data points and errors that create a complicated analysis condition, peculiarly in time series. Human interventions also affect time series, which create new patterns and further complicate the structure. In this case, neither AI nor statistical methods produce reliable results. This gap was filled by the research that the authors made in hybrid models (Zeynoddin et al. 2018, 2019; Zeynoddin and Bonakdari 2022) and it pushed the authors to go further and not only present statistical methods but also introduce one of the most recent machine learning methods, the deep learning long-short term memory model, and the hybridization theories and application.

The book is structured so an amateur reader can acquire mining and modeling knowledge by following the chapter steps. Enthusiastic readers with medium and upper knowledge of time series modeling can also appreciate the contents of the book, as the contents of MATLAB modeling are both simplified and detailed. The book presents real-world data sets and applied models in different sections for all topics. Therefore, a practical application of the methods is available, and the authors endeavored to give physical aspects to the theoretical contents—a problem that is mainly observed in statistical books.

The appendix is a well-written short course in MATLAB language. It contains all the necessary MATLAB language tutorials for the readers. It helps them save time learning the MATLAB language. With these tutorials, readers can also develop the code as they learn different applications of the functions and commands used in the book through examples and they can find corresponding necessary documents of the syntax/function in the appendix. This section is illustrative and helps the reader understand the contents well. The illustrative tutorial of the codes can also be seen in Chapters 2–5. In those chapters, the authors show how to perform a specific task by following the figures of the commands or applications of MATLAB, and they lead the reader in a very descriptive way. Offering the most primitive steps in the book, e.g., how to call data from Excel files and how to handle and model them by figures, reveals the authors’ attention to the basic needs of the readers and the authors’ intention to provide a thorough tutorial.

In addition to the simple hints, the authors also provide tutorials on customizing the other predefined functions in the book. In Chapters 2 and 3, they offer the procedure to use the online resources and obtain third-party functions’ developments that familiarize readers with other resources and how to develop and debug their codes—a simple point neglected in almost all similar books.

The chapters of the book fall into three major parts. Chapters 1–3 outline the data treatments. The acquired data sets in the real world contain many uncertainties. These uncertainties result from collection methods, their categorization, and even themselves as the appropriate model input. All these uncertainties emerge as outliers, multimodal unknown distributions, and missing points that require curating. In the first section of the book, the appropriate methods for data cleaning and treating are introduced. This section contains theoretical explanations and coding content, focusing more on coding. However, in some subsections, adding some relevant topics could improve the academic background of data, although readers could find these topics in detail in statistical books like Brandt (2014).

The second part of the book is dedicated to modeling. This section contains statistical, AI, and hybrid modeling, which are presented in Chapters 4 and 6. Different forms of stochastic models, namely, univariate and multivariate models, are shown in Chapter 4. The step-by-step introduction of models and the process of modeling procedure in MATLAB is comprehensibly defined in the chapter. The problem of the complex nature of time series and the incapability of sole modeling methods in modeling and forecasting data is addressed in Chapter 6. The authors introduce one of the most trending machine learning methods, i.e., the long short-term memory modeling method, and the procedure of its integration in the modeling process. These contents are provided without addressing the redundant and vast ranges of the AI modeling domain, making the book a ready-to-use instruction on both stochastic and deep learning methods.

The last section is devoted to the evaluation methods of the models. This section, which could be inserted as the final chapter instead of Chapter 5, defines the parsimony and diagnostic check methods, which are essential in both stochastic and AI modeling. This section of their work presents evaluation criteria and precision assessment methods that I believe suffice for model assessment. The complete examples at the end of the chapter are the strong positive points of the book. Providing instances that cover all the previous steps, including preprocessing, modeling, and evaluation, helps the reader comprehend the contents more readily.

What could be appended to the virtues? The only remark that could be counted for all chapters is some basic definitions and examples. For instance, the authors could add some examples to the deterministic and probabilistic model so the readers could better understand the underlying difference or more details in the distribution section and graphical distribution methods. The authors’ brilliant effort is that the provided YouTube tutorial covers all the book’s contents and more. The provided contents of YouTube videos contain the essential topics that were felt missing in the book. The videos also have more diverse examples, which are expected to be added to the book’s next edition. Finally, some minor typos and displacements in the book are expected to be amended in the next edition.