基于图和改进K近邻模型的高效协同过滤推荐算法

发布时间：2018-05-22 20:22

本文选题：协同过滤 + 社会网络　；参考：《计算机研究与发展》2017年07期

【摘要】：在互联网高速发展的今天,推荐系统已成为解决信息过载的有效手段,能够缓解用户在筛选感兴趣信息时的困扰,帮助用户发现有价值的信息.推荐系统中的协同过滤推荐算法,因其领域无关性及支持用户发现潜在兴趣的优点被广泛应用.由于数据的规模过大且稀疏的特点,当前协同过滤在算法实时性、推荐精确度等方面仍有较大提升空间.提出了GK-CF方法,通过建立基于图的评分数据模型,将传统的协同过滤算法与图计算及改进的KNN算法结合.通过图的消息传播及改进的相似度计算模型对用户先进行筛选再做相似度计算;以用户-项目二部图的节点结构为基础,通过图的最短路径算法进行待评分项目的快速定位.在此基础上,进一步通过并行图框架对算法进行了并行化实现及优化.在物理集群环境下进行了实验,结果表明,与已有的协同过滤算法相比,提出的GK-CF算法能够很好地提高推荐的准确度和评分预测的准确性,并具有较好的算法可扩展性和实时性能.
[Abstract]:With the rapid development of the Internet, recommendation system has become an effective means to solve information overload, which can alleviate the puzzles of users in filtering interesting information and help users find valuable information. Collaborative filtering recommendation algorithms in recommendation systems are widely used because of their domain independence and the advantages of supporting users to discover potential interests. Due to the large and sparse size of the data, there is still much room for improvement in real-time algorithm and recommendation accuracy. In this paper, GK-CF method is proposed. By establishing a graph based scoring data model, the traditional collaborative filtering algorithm is combined with graph calculation and improved KNN algorithm. Based on the node structure of the user-item bipartite graph, the shortest path algorithm of graph is used to locate the items to be graded quickly, which is based on the message propagation of graph and the improved similarity calculation model. On this basis, the parallel graph framework is used to realize and optimize the parallel algorithm. Experiments in physical cluster environment show that the proposed GK-CF algorithm can improve the accuracy of recommendation and score prediction, compared with the existing collaborative filtering algorithm. And the algorithm has good scalability and real-time performance.
【作者单位】：北京交通大学计算机与信息技术学院;北京交通大学信息中心;南京师范大学计算机科学与技术学院;
【基金】：国家重点研发计划项目(2016YFB1200100) 国家自然科学基金项目(61202429,61572256) 中央高校基本科研业务费专项资金项目(2015JBM042) 江苏省自然科学基金项目(BK20141454)~~
【分类号】：TP391.3

【参考文献】