An Effective Algorithm for Relieving Sparse Data in Collaborative Filtering Recommendation
Qin Luo-chun · 2013
Personalized recommendation system based on collaborative filtering often faces the data sparsity which seriously reduces the recommendation accuracy. An efficient hybrid weighted prediction algorithm is presented, which predicts the data visited but not rated by the characteristics and the access frequency, and fills the user-item matrix with the predictions. In this way, the sparsity of the user-item matrix is reduced, and the accuracy of recommendation is improved accordingly. Experimental results on MoiveLense data set clearly indicate that the algorithm can significantly improve the recommendation accuracy.