Abstract: Aiming at the problems that traditional collaborative filtering algorithm existed data sparseness, data redundancy and low efficiency, we proposed a collaborative filtering recommendation algorithm based on social relation and condition completion. The algorithm applied the data of social relationship into the process of matrix completion to reduce the sparse degree of the original matrix and improve the accuracy of the data completion. The vector data involved in computation was conditionally chosen to reduce the redundancy of the data and the time complexity of the algorithm in the computation of the project similarity. The experimental results show that the accuracy of recommendation of the proposed algorithm is obviously improved.

Key words: collaborative filtering, social relation, condition completion, accuracy of recommendation