Categorizing Overlapping Regions in Clustering Analysis Using Three-Way Decisions

Abstract

Clustering is a common technique for data analysis, has been widely used in many practical area. In many real applications such as social network analysis, wireless sensor networks, document clustering, and so on, there exist overlaps between different clusters due to various reasons. In this paper, we propose to use the three-way decisions approach to address categorizing overlapping regions. In contrast to existing soft clustering methods that just point out the objects whether in overlapping regions, the three-way decisions method provides a greater refinement of the categorization to system operators for further analysis, which is believed to show clearly the objects have different impacts to construct clusters. Besides, we provide a new relation-graph based clustering algorithm to obtain different overlapping region types. The results of comparison experiments are better and more reasonable to overlapping clustering.

Proximal three-way decisions: Theory and applications in social networks

“…The proposed approach to three-way 45 decision-making results from recent work on Voronoï tessellation 46 [29], Delaunay triangulation [27] of subsets in the Euclidean plane 47 and visibility in Delaunay meshes [30]. This work also carries for- 48 ward recent work on proximity (expressions of nearness between 49 sets), starting with F. Riesz (1908 [33]) and E. C ech (Brno seminar,50 May 1936 to November 1939 [38]) leading to advanced work on 51 proximity by S.A. Naimpally (from 1969 to 2014, e.g., [23,21]) 52 and others [18,19] as well as recent work on new forms of proxim-53 ity [32,28,14]. Delaunay triangulation-based proximal three-way decision-making (PTWD) 79 and rough set based three-way decision making (RTWD) models 80 are rooted in granular computing.…”

“…can be found in [44]. A more recent application in three-way deci-170 sion making clustering algorithm has been applied to a dolphin 171 social network [48]. Review of other models of set-theoretic based 172 approaches such as shadowed or interval sets are beyond the scope 173 of this paper.…”