•  
  •  
 

Applied Mathematics & Information Sciences

Author Country (or Countries)

India

Abstract

Collaborative filtering is one of the most widely used techniques for personalized recommendation services to users, since it can assist users to specify their interest on available items. The key feature of this technique is to find similar users by applying similarity measures on user-item rating matrix. Personalized system can thus provide recommendations for users based on the interest of the active user as well as a likeminded users. The success of the recommendation process depends upon the similarity metric used to find the most similar users. Similarity measures like cosine, Pearson Correlation Coefficient, Jaccard Uniform Operator Distance etc are not much effective when user-item rating matrix is sparse. This paper presents a new similarity model to calculate the similarities between each user, when only few ratings are available in the user profile. The proposed model considers both global preference as well as the local context of the user behavior. Experiments are conducted on two different datasets and compared with many existing similarity measures. The results of the experiments show that the proposed similarity measure improves the performance of the personalized recommendation process.

Digital Object Identifier (DOI)

http://dx.doi.org/10.18576/amis/110137

Share

COinS