具有用户特征约束的多关系聚类

多数聚类算法都是针对数据本身，往往忽略了用户聚类目的以及聚类过程中用户的参与指导，这样从数据本身出发的聚类结果准确性往往不太理想。针对这个问题，提出具有用户特征约束的多关系聚类算法。在多关系关联数据中进行用户参与的特征选择，用Must特征集和Can’t特征集描述用户聚类目的，通过领域本体进行特征集合扩充，得到聚类特征集合进行聚类。实验表明，该算法能较好地描述用户聚类目的，实现用户参与的聚类指导，获得了较好的聚类结果。

Multi-relational clustering with user features constraint

A lot of clustering algorithms focus on data itself.The clustering aims of users and participation,guidance of users in clustering process are neglected.It leads to inaccurate result of clustering.To solve the problem,User-Constraint Multi-Relational Clustering（UCMR-Clustering） algorithm is proposed in this paper.Features selection is guided by the user in multi-relation association data.Must-feature set and Can’t-feature set are used to describe clustering aim of the user.Features sets are expanded through domain ontology and clustering features set if acquired to cluster.The result of the experiment shows that aim of user clustering can be well described in the algorithm with user’s participation and guidance.Moreover,a good result of clustering can be obtained.