Abstract: Recommendation system is one of the most common applications in big data.Traditional collaborative filtering recommendation algorithm is directly based on user-item scoring matrix.For massive user and commodity data,the efficiency of the algorithm will be significantly reduced.Aiming at this problem,this paper proposed a collaborative filtering recommendation algorithm based on multi-relational social network.The information propagation method is used to detect communities in the multi-relationship social network based on multi-subnet composite complex network model,the users with similarity are divided into the same community.And then the k-nearest neighbor set of users is selected to construct the user-item scoring matrix within the community.Then the collaborative filtering algorithm is used to recommend through the new user-item scoring matrix,thus improving the efficiency of recommendation algorithm without reducing the accuracy of recommendation.Compared with traditional collaborative filtering recommendation algorithm on real data set Epinions,the results show that the proposed algorithm has high recommendation efficiency and accuracy.Especially for big data,the execution time of the proposed recommendation algorithm is improved by more than 10 times.

Key words: Big data, Community structure, Information propagation, Multi-subnet composite complex network model, Recommendation algorithm, Social network