A scalable collaborative recommender algorithm based on user density-based clustering
Siavash Ghodsi Moghaddam, Ali Selamat · International Conference on Data Mining · 2011
Recommender systems play an important role in online activities by making personalized recommendations to users, as finding what users are looking for among an enormous number of items in huge databases is a tedious job. The most popular recommender systems employ collaborative filtering algorithms. These methods require large amounts of training data, which cause scalability problems. One approach to solve the scalability problem is to use clustering algorithms. However, employing clustering algorithms does not always yield accurate results. We believe that by combining more accurate clustering techniques, rather than the traditional methods, with collaborative filtering algorithms, the accuracy and scalability of the recommender system will be improved. In this paper we propose a hybrid recommender system, which is composed of a density-based user clustering method based on users' demographic information and user-based collaborative filtering. Experiments have been conducted to evaluate our approach using MovieLens dataset. The experimental results have shown that the proposed method improves accuracy as well as scalability.