A New Approach for Recommendation System Using Improved User Similarity Matrix

Abstract

Recommendations systems are very useful in recommending items for online portals. These systems are applicable in e-commerce web sites, publication agencies, news agencies etc. Many researchers proposed recommendation algorithms on the basis of items similarity, users similarity or hybrid of both. This paper proposed a recommendation system that is based on the history of users. It finds the similarity between users by exploring their previous history. The algorithm is applied on a standard data set of movie recommendation and its results are found better as compared to other algorithms.