An Improved Item-based Collaborative Filtering Algorithm Based on Clustering Method

Item-based collaborative filtering recommendation algorithm is one of the most widely used recommendation algorithm, which is widely used in many recommendation systems. But there are some drawbacks when used in large e-business systems. The existing traditional algorithms can’t perform well when the item space changes; on the other side, the performance of the recommendation system will go down as the items increase into a large amount. An improved collaborative filtering recommendation algorithm based on dynamic item clustering method was proposed in the paper. A similitude threshold model was introduced to divide the item space into clusters dynamically. Experiment states that using dynamic item clustering method can satisfy the requirement of increasing amount users and consumers in large e-business systems. The improved collaborative filtering recommendation algorithm performs relatively good recommendation with less resource consumption.