Collaborative Recommendation Systems and Link Analysis
Link analysis is a framework usually associated with fields such as graph mining, relational learning, Web mining, text mining, hyper-text mining, visualization of link structures. It provides and analyzes relationships and associations between many objects of various types that are not apparent from isolated pieces of information. This chapter shows how to apply various link-analysis algorithms exploiting the graph structure of databases on collaborative-recommendation tasks. More precisely, two kinds of link-analysis algorithms are applied to recommend items to users: random-walk based models and kernel-based models. These link-analysis based algorithms do not use any feature of the items in order to compute the recommendations, they first compute a matrix containing the links between persons and items, and then derive recommendations from this matrix or part of it.