基于节点重要度的用户推荐

为用户推荐具有相同兴趣的好友是用户关系的研究热点之一，但面临着数据稀疏性、兴趣推荐偏差等问题。针对这些问题，本文提出一种考虑节点重要度的用户推荐方法。首先结合共现及凝聚方法实现标签聚类，据此划分具有相似兴趣的用户社群。然后通过社会网络分析构建社群的用户关系网络，采用PageRank计算用户重要度指标用于评价用户的推荐能力，并描述了用户推荐算法流程。最后通过在真实数据集上与传统方法的对比实验，验证了模型的有效性，给出了表示用户亲疏关系的可视化结果。

User Recommendation Based on Node Importance Degree

Recommending friends with the same interest to a user is a hot issue of user relationship research. But it also has many problems, such as data sparseness and recommendation deviation in interest. To address such problems, this paper proposed a recommender approach considering node importance degree (NID). First, through the combination of co-occurrence and agglomerate approaches, it realized a tag clustering algorithm and divided user groups with similar interest. Second, it constructed user relationship network through social network analysis. Based on it, NID in PageRank was calculated to evaluate user recommended ability. Compared with several other approaches on real datasets, the approach achieved the best performance. We finally also presented visualized results to express user relationship.