Active selection of clustering constraints: a sequential approach

This paper examines active selection of clustering constraints, which has become a topic of significant interest in constrained clustering. Active selection of clustering constraints, which is known as minimizing the cost of acquiring constraints, also includes quantifying utility of a given constraint set. A sequential method is proposed in this paper to select the most beneficial set of constraints actively. The proposed method uses information of boundary points and transition regions extracted by data description methods to introduce a utility measure for constraints. Since previously selected constraints affect the utility of remaining candidate constraints, a method is proposed to update the utility of remaining constraints after selection of each constraint. Experiments carried out on synthetic and real datasets show that the proposed method improves the accuracy of clustering while reducing human interaction.