关键词: 主动学习, 聚类算法, 动态模糊集, 动态模糊关系, 动态模糊测度

Abstract: Data clustering has recently become a topic of significant interest to data mining and machine learning communities. Because achieving supervised data may be expensive, the research focuses on attaining the supervised information form little information but can significantly improve the clustering performance. Moreover, there are many dynamic fuzzy problems in the real world. This paper presents a dynamic fuzzy data clustering algorithm based on active learning, and introduces three constraints which include dynamic fuzzy equivalence relation, dynamic fuzzy trust measure and dynamic fuzzy likelihood measure to guide the clustering process of DBSCAN, aiming at improving clustering performance. Experimental results show that this proposed approach is effective in data clustering; also it can describe the dynamic fuzzy data of the clustering problem better. The clustering performance of active DF-DBSCAN has been dramatically improved with three constraints and better than the three representative methods.

Key words:  active learning, clustering algorithm, dynamic fuzzy sets, dynamic fuzzy relation, dynamic fuzzy measure