Search

Recommender Systems are useful information filtering tools that have reduced information overload over the web. Collaborative filtering (CF) is one of the extensively used recommendation techniques. Traditional CF captures user-item ratings in a two-dimensional rating matrix which does not sufficiently convey user preferences. Ratings based on several criteria are incorporated into CF to develop multi-criteria recommender systems (MCRS). MCRS are more efficient and cater to the users’ needs with more satisfaction. However, there are certain issues like multidimensionality, sparsity, and cold start associated with MCRS. This paper aims to study the MCRS and investigate efficient solutions for existing issues. In this direction, we proposed a modified similarity measure that improves the accuracy of neighborhood generation and rating prediction. In the proposed approach, the users are clustered based on multi-criteria ratings, which reduces the data sparsity and multidimensionality issues in MCRS. The supremacy of the proposed approach is verified by conducting experiments on a benchmark data set and evaluating the performance using some standard evaluation measures.