Tag-cloud based recommendation for movies

Abstract

Most of the recommendation systems aim to make suggestions for individuals rather than a group of users. However, people are sociable and most of the items to be recommended like movies, restaurants, tourist destinations, etc. are for group consumption. Making recommendations for a group is not a trivial task due to the diverse and conflicting interests of the group members. In this paper, we present a framework for recommending movies to a group of users. Existing recommendation systems for movies use users’ ratings as a measure to suggest individual recommendations or use them to generate the group profile by using aggregation methods in case of group recommendations. In this work, we focus on two things: exploiting the tags assigned to the movies by the users and leveraging the semantic information present in them to make recommendations. The assigned tags along with their weightages are used to form tag clouds for individual group members as well as for movies. Following this a Group Score is computed for each movie on the basis of the content similarity of the tag cloud of the group and the tag cloud of the movie. The movies having top-N Group Score are recommended to the group. To verify the effectiveness of this framework, experiments have been conducted on the MovieLens-10M and MovieLens-20M datasets. Results obtained clearly demonstrate how the accuracy of the recommendations increase with the increase in the homogeneity of preferences within the group members.