English

Abstract

In recent years Internet has emerged as an important tool for gaining information. But people confronts with abundance of data. High performance information searching is required to cope with this situation. Recommender systems search appropriate documents or filter out inappropriate documents from several information streams in order to match with a user's general interests. Recent developments in recommender systems implements user classes as per the demographic data. Collaborative Filtering is the most vital component of recommender system as it recommends items by considering the ratings of similar users. We argued that additional factors have an important role to play in guiding recommendation. In this paper, we propose the design of Credit Based Collaborative Filtering (CBCF) approach for recommender systems which employs prioritized features of items to improve the efficiency of recommendations. In our suggested plan, each recommender is assigned with a credit value which signifies the goodness of a recommender. The profile similarity in combination with credit value influences the decision making ability of recommender system. Further this credit value gets updated after every recommendation as user gives feedback about likeliness of the item. Eventually these modified credit values make the recommender system learn and thereby to improve the prediction accuracy in comparison to the classical recommender systems. We have supplemented our approach with a case study of a movie recommender system. A comparison of the generated recommendations using CBCF approach to the classical approach establishes the validity of the proposed system.