A collaborative filtering recommendation algorithm based on the influence sets of e-learning group’s behavior

At present, due to use of nearest neighbor query algorithm based on memory, the traditional user-based collaborative filtering (CF) recommendation system has the shortages of poor expandability and lack of stability. On the aspect of expandability, the item-based CF algorithm was proposed but it still has not addressed the declined recommendation quality (poor stability) caused by sparse data. Being inspired by the concept of influence set, this paper proposes a new recommendation algorithm of CF-ISEGB (collaborative filtering based on the influence sets of e-learning group’s behavior). The influence sets of current e-learning group are used to improve the evaluation density of this resource, and the computation, prediction and rating methods are also defined for this new recommendation mechanism. The experimental result shows that compared to the traditional item-based CF algorithm which generates recommendation only based on the nearest neighbor, it can effectively alleviate the problem caused by sparse datasets and significantly improve the recommendation quality of recommendation system.