An Improved Similarity Metric for Recommender Systems

Abstract: Due to pervasive technologies in various applications, which are used in our everyday lives, recommender systems have become widely used in most of these applications to estimate the users’ needs depending on his/her preferences. The development of recommendation methods typically focuses on maximizing the prediction accuracy of the users’ interests. Currently, collaborative filtering (CF) is a widely used approach for recommender systems. The similarity measures play a major role in such recommender systems. In spite of the availability of many different similarity measures, user similarity is yet to be calculated perfectly in recommender systems. We propose a similarity metric that helps to increase the accuracy of recommended items.