基于多元关系的张量分解标签推荐方法

标签推荐的现有方法忽视了多种属性特征之间的联系，无法保证大数据环境下推荐系统的准确率。针对该问题，提出了一种基于用户聚类和张量分解的新标签推荐方法，以进一步提高标签推荐的质量。该方法首先对一些对产品具有重要影响的用户进行聚类，然后根据用户、产品、标签和产品评分之间的多元关系综合计算总权重。最后，根据聚类之后的用户群体以及多元关系的总权值构建张量并进行张量因式分解。实验与传统张量分解方法相对比，结果表明提出的方法在准确率上具有一定的提高，验证了算法的有效性。

Method for tag recommendation of tensor decomposition based on multiple relationships

The exist method of tag recommendation ignore the connection among the characteristics of a variety of attributes and can not guarantee the accuracy of the recommender system in the big data environment. Aiming at this problem, this paper proposed a tag recommendation method based on user clustering and tensor decomposition, which could further improve the quality of tag recommendation. The method firstly clustered the users who had an important influence on the product, and then comprehensively calculated the total weight based on the multiple relationships among the users, products, tags, and product ratings. Finally, it constructed the tensor according to the user groups after clustering and the total weight of the multivariate relations, and performed the tensor factorization. Experiment compared with the traditional tensor decomposition method, and the results show that propsed method improves the accuracy and verifies the effectiveness of the algorithm.