关键词: 矩阵分解, 聚类, 协同过滤, 推荐准确性

Abstract: Based on the matrix decomposition and clustering, we proposed a collaborative filtering recommendation algorithm. The algorithm first used alternating least squares (ALS) algorithm to decompose matrix, and then used improved k-means clustering algorithm to compensate for large amount of calculation caused by single ALS algorithm in the later stage of collaborative filtering. The problem that the recommendation accuracy was low due to the reduction of the high dimension and high sparsity of original matrix was solved, and the calculation speed and recommendation accuracy were greatly improved. Experimental results show that the improved algorithm significantly improves the recommendation accuracy.

Key words: matrix decomposition, clustering, collaborative filtering, recommendation accuracy