Abstract: The traditional recommendation system usually uses collaborate filtering or content-based recommendation as its method,but this paper applied matrix completion.Because of the sparsity of data,if matrix completion is used directly,error will be relatively large.Considering that some active users exist in all users,by means of finding these special users,the sparsity will be reduced by their integrated data.And improving recommendation quality for these users will be more likely to generate values.This paper proposed a split-integration recommendation algorithm,and the experimental results show that the proposed method can improve the accuracy of recommendation.

Key words: Recommendation systems,Matrix completion,Active users,Split-Integration