Realtime data mining. Self-learning techniques for recommendation engines (Q2393375)

The book is based on the assumption that real-time recommendation engines (e.g. for a web-shop) should use the framework of reinforcement learning (RL) and Markow decision processes (MDPs): ``In the simplest case, the product detail views form the states, the recommended products the actions, and the rewards the clicks or purchases of the products. The goal consists (depending on the chosen reward) in maximizing the activity (clicks) or the success (sales) \dots the recommendation engine is following the proven recommendations (exploitation)\dots (or) \dots a new recommendation is tried out (exploitation).'' (p. 41). The time-based transaction behavioral data (clicks, sales) are furthermore simplified by matrix (Chapter 8) and tensor (Chapter 9) factorization methods. A combined RL/tensor factorization engine is proposed in Chapter 10, statistical model evaluation tests in Chapter 11, and the partial open source software library XELOPES in Chapter 12. This library represents the implementation of the proprietary recommendation engine of the PRUDSYS AG.