Exploiting formal concept analysis in a customizing recommendation for new user and gray sheep problems

Recommender systems are becoming an indispensable application and re-shaping the world in e-commerce scopes. This paper reviews the major problems in the existing recommender systems and presents a tracking recommendation approach based on information of user's behavior and two-level property of items. A new recommendation model based the synergistic use of knowledge from repository, which includes user's behavior, and items property was constructed and utilizes the Formal Concept Analysis (FCA) mapping to guide a personalized recommendation for user. We simulate a prototype recommender system that can make the quality recommendation by tracking user's behavior for implementing the proposed approach and testing its performance. Experiments using two datasets show our strategy was more robust against the drawbacks and preponderate over traditional recommendation approaches in cold-start conditions.