Document Type



Collaborative filtering uses information about customers’ preferences to make personal product recommendations and is achieving widespread success in e-Commerce. However, the traditional collaborative filtering algorithms do not response accurately to customers’ needs. The quality of the recommendation needs to be improved in order to support personalized service to each customer. In this paper, we present novel method to improve the accuracy of the collaborative filtering algorithm. We borrow vector space model from information retrieval theory and use it to effectively discriminate the preference weights on the items for each customer. The proposed method achieves more accurate recommendations for customers who purchase similar types of products repeatedly. Our experimental evaluation on the well-known MovieLens data set shows that our method does result in a better accuracy.