Comprehensive Diversity in Recommender Systems

Abstract

The present age of digital information has presented a heterogeneous online environment which makes it a formidable mission for a noble user to search and locate the required online resources timely. Recommender systems were implemented to rescue this information overload issue. However, majority of recommender systems focus on the accuracy of the recommendations, leaving out other important aspects in the definition of good recommendation such as diversity and serendipity. This results in low coverage and long-tail items are often left out in the recommendations. In this paper, we present and explore a recommendation technique that ensures that comprehensive diversity is also factored-in in the recommendations. The algorithm adopts the second line of recommendation improvement whereby a recommendation list is re-ranked in such a way that it would include long-tail items. The results showed that the proposed algorithm is capable of giving a balanced list of recommendations in terms of accuracy and diversity.