Overlapping Granular Clustering: Application in Fuzzy Rule-Based Classification

• Published in: RECENT ADVANCES ON SOFT COMPUTING AND DATA MINING, SCDM 2024
• Main Authors: Muda, Muhammad Zaiyad; Panoutsos, George
• Format: Proceedings Paper
• Language: English
• Published: SPRINGER INTERNATIONAL PUBLISHING AG 2024
• Subjects: Computer Science
• Online Access: https://www-webofscience-com.uitm.idm.oclc.org/wos/woscc/full-record/WOS:001324620600009

A clustering technique often aims to create a number of disjoint clusters or granules, in which an element or instance is only permitted to belong to one cluster. However, the majority of real-world data sets have information that overlaps, causing specific data objects or patterns to belong to multiple clusters. For instance, an individual may concurrently be a member of more than one social group, such as a family group and a friend group. Therefore, the purpose of this study is to use a parameter called R-value to allow and provide parametric control for cluster overlaps. In this research, it is demonstrated that the inclusion of R-value in the Granular Clustering (GrC) enables GrC to control the amount of overlapping between clusters. Datasets from the UCI Machine Learning Repository are used to illustrate the new GrC algorithm with overlapping measure. Results reveal that the GrC with overlapping measure surpasses the traditional GrC in terms of classification accuracy, highlighting the possible application of the overlapping GrC for creating Fuzzy Logic rule bases.