基于异构信息网络的模糊贴近度推荐算法

针对传统协同过滤算法在用户推荐过程中数据稀疏性、可扩展性、用户兴趣迁移变化等问题,提出一种基于异构信息网络的模糊贴近度推荐算法。在k-means聚类算法基础上构建新的异构信息网络,利用关系抽取的方式构造用户属性权重矩阵;引入模糊贴近度综合分析元路径属性权重的影响,寻找近邻用户;采用Top-N算法排序进而完成推荐,并进行准确性验证。在Epinions数据集上的实验结果表明,在推荐质量和速度上,所提推荐算法较传统推荐算法更优。

Fuzzy nearness recommendation algorithm based on heterogeneous information network

Aiming at the problems of data sparsity,scalability and user interest migration in the process of user recommendation in the traditional collaborative filtering algorithm,a fuzzy closeness recommendation algorithm based on heterogeneous information network was proposed.A new heterogeneous information network was constructed based on the k-means clustering algorithm and the user attribute weight matrix was constructed using the relationship extraction method.The fuzzy closeness was introduced to comprehensively consider the influence of the weight attribute of the meta path and find the neighboring users.The Top-N algorithm was adopted to sort out and then complete the recommendation and verify the accuracy.Experimental results on the Epinions dataset show that using the proposed recommendation algorithm improves the recommendation quality and speed compared with using the traditional recommendation algorithm.