基于用户近邻约束的矩阵因子分解算法

摘要

Matrix factorization algorithm based on collaborative filtering is one of the most widely used in the personalized recommenda-tion system. Concerning the problems of data sparsity and real-time in recommendation system,a matrix factorization algorithm based on user’ s neighbors regularized is proposed based on traditional matrix factorization model. The algorithm takes advantage of the matrix fac-torization model,using the user’ s neighbor as a regularization to improve the quality and real-time of recommendation algorithm. The ex-perimental results in movieLens datasets show that the proposed algorithm can more efficiently improve recommendation quality than the traditional algorithm,and solve the problems of data sparsity and real-time.%矩阵因子分解推荐算法是基于模型的协同过滤算法中应用最广泛的一种推荐技术。针对推荐系统数据的稀疏性和推荐算法的实时性等问题，在传统矩阵因子分解模型的基础上引入用户近邻模型约束，提出基于用户近邻约束的矩阵因子算法。该算法充分利用了矩阵因子模型的优点，通过用户近邻约束进一步提高了算法相应的实时性和推荐的质量。在MovieLens数据集上的实验结果表明，该算法能有效解决数据稀疏和实时性问题，在推荐质量上比传统算法有了较大提高。