The general scope of fraud analytics comprises a set of both conceptual approaches and appropriate tools for detecting certain anomalies in processes that can be associated with fraud. Fraud can occur in several business domains, spanning from banking to healthcare to telecom. Fraudsters are always looking for niches allowing them to earn revenue by carefully organizing a concealed crime. The common panacea to all these frauds is efficient data analytics, which can capture irregularities either when comparing overall behavior case by case, or by detecting subtle time-varying properties for individual cases.

The book is structured in seven chapters and provides an excellent tradeoff between technical chapters, addressing machine learning and graph theoretical modeling, and the more process-driven chapters. This makes this book unique; although other books covering fraud analytics exist (among which I could highlight [1]), this book by Baesens et al. manages to show both viewpoints. The first viewpoint on fraud analytics is given in chapters 1, 6, and 7. In the first chapter, the authors give some concrete examples on types of fraud (insurance, credit cards) and their estimated financial impact. The last two chapters (6 and 7) delve into how fraud analytics can leverage post-processing and benefit from data quality management. Back-testing, visual analytics, and economic issues related to the process itself are discussed thoroughly in the last part of the book.

I was fascinated, however, by the four main technical chapters (2 through 5), representing an incredible richness of knowledge on exploratory data analysis (chapter 2), clustering (chapter 3), supervised machine learning (chapter 4), and social networking (chapter 5).

Besides the excellent coverage and illustrations, the authors provide many additional references to updated and relevant academic sources. The logical flow throughout the book is outstanding, and each chapter on its own can serve as mandatory background reading for graduate students, researchers, or applied data analysts. In fact, the technical content is so well written that the book might be appropriate even for a general introduction to data analysis having fraud analytics only as an application domain.

For these reasons, I recommend the book to any reader interested in applied data science or fraud analytics. It is highly relevant and excellent both in content and form--definitely a must-read in this domain.

More reviews about this item: Amazon