A New Multi-criteria Recommendation Algorithm for Groups of Users

摘要

Group recommendation algorithms have the advantage of helping groups of users find favorite items under less time. And, the multi-criteria recommendation system aims at obtaining the user's preferences over various aspects and make accurate recommendation. This paper presents a new group-oriented multi-criteria recommendation algorithm called GMURec. This algorithm first uses K-means algorithm which generates the groups (i.e., sets of users who have similar interests). Then, it uses the BP neural network to aggregate the groups preferences and learn the implicit relationship between multi-ratings and overall rating of each group. The performance of GMURec algorithm is compared with three baseline algorithms. Experimental results show that: (1) the precision of GMURec is only lower than the individual-targeted personal multi-criteria recommendation algorithm, but is higher than the other two group ones; (2) the recall of GMURec is as good as or better than other algorithms; (3) the run time of GMURec is the least one among the compared algorithms.