Three-Way Clustering: An Advanced Soft Clustering Approach

Clustering is a machine learning technique that assigns unlabelled data points into different groups based on similarity of data. However, in many cases, we are unable to confidently assign data points to particular clusters. Soft clustering introduces a probability of the data points belonging to different clusters. Three-way clustering is a recent development of soft clustering based on three-way decisions. In particular, each data point is assigned a value to represent if it is inside, outside, or partially inside a cluster. There are two types of three-way clustering techniques, namely, evaluation-based approaches and operation-based approaches. The evaluation-based approaches rely on a membership function to calculate the degree of a data point belonging to a cluster. The operator-based approaches use a pair of operators to construct a three-way cluster from a hard two-way cluster. We will introduce, review, and analyse various three-way clustering techniques in this paper. In addition, the history of three-way clustering and the future development of three-way clustering will also be discussed.