Frequency-based similarity measure for context-aware recommender systems

Abstract

Collaborative Filtering (CF), the widely used and most successful technique in the area of Recommender Systems, provides useful recommendations to users based on their similar users. Computing similarity among the users efficiently is the major step in CF. Further, it has been observed from literature that the context into CF provides more accurate and relevant recommendations for users but it is hard to represent and model contextual factors directly into the system. In this paper, we have incorporated the contextual information into user profile as an additional feature through a proposed novel frequency count method. After extending the user profiles, items are recommended based on similar profiles computed through a novel similarity measure. To evaluate the performance of our proposed recommendation strategy, several experiments are conducted on the popular LDOS-CoMoDa dataset.