Integration of users preferences and semantic structure to solve the cold start problem

Abstract

Web recommendation systems aim to find the most interesting and valuable information for web users based on their collected preferences. Although, the collaborative filtering approach is the widely used, but it suffers from several problems, one of these problems is known as the cold start problem (for example, if a new user visit Amazon web site for first time, then the Amazon system becomes unable to generate recommendations). We suggested the active node technique as a method of solution to the cold start problem, and we integrate collected users' preferences within a semantic structure, and we compare between non-semantic and semantic structure of the active node method based on three criteria novelty, coverage, and precision of generated recommendations. We found that the semantic structure achieve higher performance than non-semantic.