Algorithms for Behavior-Based Recommender Systems

Different methodologies exist for the generation of behavior-based recommendations. This chapter focuses on methods that are especially suited to the needs and challenges associated with the application area of STI providers. The behavioral input data consists of market baskets that can be found likewise in e-commerce, library environments, or (Web 2.0) social network sites. The relevant problem that has to be solved is the question, which co-purchases or co-inspections of products in the market baskets occur non-randomly thus hinting at an underlying relation of these products. First, a method developed for large samples based on calculating inspection frequency distribution functions following a logarithmic series distribution is presented.