基于用户聚类和移动上下文的矩阵分解推荐算法研究

Study on Matrix Factorization Recommendation Algorithm Based on User Clustering and Mobile Context

随着移动互联网技术的快速发展,越来越多的用户通过移动设备获取移动信息和服务,导致信息过载问题日益凸出。针对目前上下文感知推荐算法中存在的数据稀疏性差、上下文信息融入不够、用户相似性度量被忽略等问题,提出一种基于用户聚类和移动上下文的矩阵分解推荐算法。该算法通过利用k-means对用户聚类找到偏好相似的用户簇,求出每簇中并对用户所处上下文之间的相似度并对其进行排序,由此找出与目标用户偏好和上下文均相似的用户集合,借助该集合改进传统矩阵分解模型损失函数,并以此为基准进行评分预测和推荐。仿真实验结果表明,所提算法可有效提高预测评分的准确度。

With the rapid development of mobile Internet technology,more and more individuals use mobile devices to acquire information and services,which makes information overload problem more and more serious.Aiming at the puzzle resulted from data sparsity,insufficient contextual information and ignoring context similarity measurement,this paper proposesd a method of matrix factorization recommendation algorithm based on user clustering and mobile context(UCMC-MF)to predict user ratings and make recommendation.Firstly,the method clusters similar user by way of k-means,then finds similar contexts in each cluster,and searches users who are similar to the target user in preferences and context.Finally,experimental results on real datasets demonstrate that the proposed algorithm can effectively improve the accuracy of prediction.