A Re-ranking Technique for Diversified Recommendations

User satisfaction is the most important challenge for any user oriented system. Especially in today's world where tremendous amount of information is available, which can be used for knowledge discovery to find out user's interest. Recommender systems which are simulations of web personalization are now days widely integrated in various domains for quality improvements. Recent studies has shown that to improve user satisfaction one should also consider other quality factors such as diversity rather than relying only on accuracy of recommendations. We propose a hybrid approach of recommendation which re-ranks the most relevant predicted items according to the specified criteria MCBRT. We aim at maintaining substantially higher aggregate diversity across all users while maintaining adequate level of recommendation accuracy.