Information overload has been a known phenomenon for at least the last two decades. Reordering or filtering search results, individualizing searched information, and providing recommendations are essential, well-known techniques to reduce information overload. Recommendation is the main topic of this book, but the title suggests a different main topic: machine learning. In fact, the book interconnects recommendation and machine learning with a principal emphasis on recommendation. Because recommenders have been with us for many years, I understand the book title as an ambition to distinguish it and highlight a hot present-day topic: machine learning. However, it can be a bit misleading to prospective readers (the subtitle “Applications in recommender systems” is easily overlooked on the book cover).

Researchers dealing with problems of accessing high volumes of complex data will make the best use of this book. Even though it is primarily a research text, the authors extensively present existing approaches to recommender systems and machine learning in a tutorial style. Therefore, graduate students can also use the book to get an overview of recommender systems and selected machine learning approaches. They will also learn about the application of some concepts in a specific domain (music).

Lampropoulos and Tsihrintzis follow a typical research text structure: introduction, state-of-the-art, proposal description, evaluation, conclusions, and future work. After an introductory chapter, the authors present in chapter 2 a review of previous work related to recommender systems. These 15 pages and 53 references can be well used by a novice in recommender systems research. Chapter 3 introduces selected topics from the machine learning domain, namely supervised learning, more specifically the problem of classification and the support vector machine as a method for supervised learning, which is often utilized in recommender systems.

Chapter 4 provides necessary background for content-based recommendation in the music domain. It presents the main multimedia data descriptors, which are used in chapter 5, proposing the content-based recommender system called MUSIPER. MUSIPER combines certain subsets of objective (content) features and the subjective music similarity perception of individuals acquired by neural network-based incremental learning. I found this part to be rather interesting and novel. It is a pity that it is described without a comprehensive background.

The main contribution of the authors’ research is described in chapter 6 and evaluated in chapter 7. The hybrid recommender proposed in the book utilizes only positive examples given by users (the examples can also be inferred from the users’ behavior). The recommender involves two levels of classification of items: the first level involves a one-class classifier trained exclusively on positive examples, and the second level is realized in two alternatives as a multiclass classifier or collaborative filtering.

Throughout the book, the authors give the formalization of presented problems with references to principal works in the domain of recommender systems and machine learning. However, the authors do not go beyond this. I would expect more discussion on recent advancements and a broader view in the summary parts (there are no recent references: the latest was published seven years prior to this book’s publication). The conclusions chapter just repeats on one page what is already presented in the book (in the foreword, preface, and introduction). Two paragraphs on future work deal only with specific improvements. In spite of such a narrow view, I will recommend the book to my graduate students as a nice piece of research including well-presented background and good evaluation methodology.