Community Detection and Disease identification using Meta-heuristic based Clustering methods

Abstract

Community detection in the field of medical science is an important field of research for disease diagnosis. A popular data mining technique for community detection is data clustering which partitions huge volume of data items based on some similarity measures. This paper presents a rigorous data clustering application with its practical implications towards medical field of research for disease diagnosis. The comparison study involves six benchmark UCI machine learning medical datasets and four meta-heuristic based partitional clustering algorithms such as Sine Cosine Algorithm (SCA), Grey Wolf Optimization (GWO), JAYA Algorithm and Particle Swarm Optimization (PSO) based partitional clustering method. To statistically verify the statistical significance of each model, the performance of each model, is accessed using Duncan’s multiple range test on each individual dataset separately considering four performance measures such as Accuracy, F-score, MCC and Kappa. In addition, we have conducted Nymenyi hypothesis test to obtain the rank of each model considering the results of all datasets altogether. The results obtained from the statistical test signifies that the SCA algorithm obtains statistically superior performance among the remaining algorithms considered in this study.