A Method to Maintain Item Recommendation Equality Among Equivalent Items in Recommender Systems

Abstract

Collaborative Filtering is a useful algorithm to offer personalized recommendations for users. However, there are several technical challenges in collaborative filtering, including the first-rater problem, where an item not yet evaluated cannot be recommended until it has been rated. In the paper, the presenting method deals with the first-rater problem that is similar to the process starvation is operating systems. The method reduces the score gap between items and makes it possible for a new item or an item with no user preference to be recommended automatically. Thus, the system can recommend items in the same group without bias. Finally, we present an analysis of an example of the algorithm.