New Book on Realtime Analytics and Recommendation Engines
The book covers realtime analytics and its application to recommendation engines from a control-theoretic perspective.
By Alexander Paprotny and Michael Thess.
Springer, 2013, 313 pages.
Describing novel mathematical concepts for recommendation engines, this book features a sound mathematical framework unifying approaches based on control and learning theories, tensor factorization, and hierarchical methods. Furthermore, it presents promising results of numerous experiments on real-world data.
This monograph appeals to computer scientists and specialists in machine learning, especially from the area of recommender systems, because it conveys a new way of realtime thinking by considering recommendation tasks as control-theoretic problems. The book will also interest application-oriented mathematicians because it consistently combines some modern mathematical areas, especially from approximation theory.
The main goal of the research presented in the book consists of devising a sound and effective mathematical and computational framework for automatic adaptive recommendation engines. Most importantly, the authors introduce an altogether novel control-theoretic approach to recommendation based on considering the customer of an (online) shop as a dynamic system upon which the recommendation engine acts as a closed-loop control system, the objective of which is maximizing the incurred reward (e.g., revenue).
This book also covers classical data mining based approaches and develop efficient numerical procedures for computing and, especially, updating the underlying matrix and tensor decompositions. Furthermore, the authors take a step toward a framework that unifies the two approaches, that is, the classical and the control-theoretic one. In summary, the book proposes a very modern approach to realtime analytics and includes a lot of new material.