Resolving the Challenges of Partition Based Clustering Methods Using Hybrid PSO _SGO and Simrank Ensemble Methods

Abstract

Identifying meaningful clusters in data is one of the most important goals of newlineunsupervised learning. Determining clusters of data points, or clusters, that fit together newlinebecause they are related in some way is the goal of clustering algorithms. Numerous newlinepapers on clustering-related research have revealed that clustering problems are not newlinewithout their difficulties. Even if several scientists have advocated different persistence newlinestrategies for many years, there still appears to be a requirement for the auxiliary newlineextension of those strategies. In this research, the study is limited to the K-means newlinealgorithm and the partitioned clustering algorithm in general. To tackle the problems newlinewith K-means, there are some significant and well-known hurdles. The primary goal of newlinethis work is to investigate and analyse the traits, difficulties, and performance issues of newlineconventional and evolutionary clustering methods using established validation criteria. newlineThese partition-based clustering models do not have adequate global optimal newlineconvergence. Partition-based clustering techniques are used to create evolutionary newlineclustering models, which address these problems. These techniques offer a better newlinesolution to the problems than partition-based clustering. Even though certain newlineevolutionary theories have had issues with the emergence of new clusters, individuals newlineare unable to fix the local optimization issues. Therefore, it was proposed to use hybrid newlinePSO and SGO evolutionary algorithms to solve local optimization problems. newlineAdditionally, it increases the clusters rate of convergence. The application of K-means newlinewith a cluster guess value in this hybridization will not yield the desired outcomes.Stochastic processes underlie soft computing techniques. newlineThese models feature a random probability distribution that can be newlinequantitatively analysed but not accurately predicted. Because it increases the newlineeffectiveness of the cluster results, cluster ensemble has become a popular technique newlinefor cluster analysis. newline