Kernel-based exponential grey wolf optimizer for rapid centroid estimation in data clustering

Abstract

Clustering finds variety of application in a wide range of disciplines because it is mostly helpful for grouping of similar data objects together. Due to the wide applicability, different algorithms have been presented in the literature for segmenting large multidimensional data into discernible representative clusters. Accordingly, in this paper, Kernel-based exponential grey wolf optimizer (KEGWO) is developed for rapid centroid estimation in data clustering. Here, KEGWO is newly proposed to search the cluster centroids with a new objective evaluation which considered two parameters called logarithmic kernel function and distance difference between two top clusters. Based on the new objective function and the modified KEGWO algorithm, centroids are encoded as position vectors and the optimal location is found for the final clustering. The proposed KEGWO algorithm is evaluated with banknote authentication Data Set, iris dataset and wine dataset using four metrics such as, Mean Square Error, F-measure, Rand co-efficient and jaccord coefficient. From the outcome, we proved that the proposed KEGWO algorithm outperformed the existing algorithms.