The important phase in ensemble clustering is the consensus function. In terms of what is the goal for comparison in the consensus process, this study divides all consensus functions into four categories: partition-partition (P-P) comparison, cluster-cluster (C-C) comparison, member-in-cluster (MIC) voting, and member-member (M-M) co-occurrence. Each ensemble clustering approach is divided into two steps: generation and consensus. P-P comparison approaches, also known as median partition approaches, aim to solve an optimization problem by maximizing the total similarity to the specified partitions. C-C Comparison treats a cluster as a unit and examines cluster similarity. MIC Voting is a concept shared by several systems that focus on the relationship among members and their clusters across all partitions. The M-M co-occurrence method transforms the consensus partitioning problem into a co-association matrix partitioned problem.

Consensus Function; Clustering; Voting; Co-Occurrence; Ensemble