An integer programming clustering approach with application to recommendation systems

Recommendation systems have become an important research area. Early recommendation systems were based on collaborative filtering, which uses the principle that if two people enjoy the same product they are likely to have common favorites. We present an alternative recommendation approach based on finding clusters of similar customers using integer programming model which is to find the minimal number of clusters subjected to several similarity measures. The proposed recommendation method is compared with collaborative filtering, and the experimental results show that it provides relatively high prediction accuracy as well as relatively small variance.