An Improvement to Collaborative Filtering for Recommender Systems
Li-Tung Weng, Yue Xu, Yuefeng Li, Richi Nayak · 2006
Collaborative filtering recommenders utilize a database of user preferences to make personal product suggestions, and have achieved widespread successes in various e-commerce applications nowadays. Inverse user frequency is one of most well known approaches to improve the accuracy of the standard collaborative filtering recommender. In this paper, we propose a statistical attribute distance method that uses the similarity in statistics of users' ratings to calculate the user correlation instead of using the statistics of users that rate for similar items. Form our experiment results we suggest the statistical attribute distance outperforms inverse user frequency in recommendation accuracy and scalability