Title: Improving the accuracy of item recommendations in collaborative filtering using time-variant system

Authors: Desabandhu Parasuraman; Sathiyamoorthy Elumalai

Addresses: School of Information Technology and Engineering, VIT University, Vellore 600014, India ' School of Information Technology and Engineering, VIT University, Vellore 600014, India

Abstract: Nowadays, the real challenge is to manage the dynamic web content in order to provide a prolific resource to the user. Web personalisation is an outcome of the challenge by which the web is a tailor made to a user. Recommendation systems access the user profile using collaborative filtering and content based filtering to provide better personalisation. This paper focuses on improving the accuracy of item recommendations, based on the dynamic item-based collaborative filtering by utilising time variant system which is implied on user ratings. Similarity between the items is found by using vector similarity and weight is calculated by Pearson correlation coefficient. Comparison of the results of traditional item-based collaborative filtering with dynamic item-based collaborative filtering is also discussed. Finally, it is observed that the user's dynamic voting average improves the accuracy of recommendations comparing to the normal voting average on items.

Keywords: recommendation systems (RS); user profiles; collaborative filtering (CF); time variant system; vector similarity.

DOI: 10.1504/EG.2017.087948

Electronic Government, an International Journal, 2017 Vol.13 No.4, pp.324 - 338

Received: 29 Dec 2016
Accepted: 13 Feb 2017
Published online: 13 Nov 2017 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article