STUDY OF CENTROID INITIALIZATION IN K MEANS CLUSTERING ALGORITHM

Abstract

K-mean Clustering is a feature-based feature detection and similarity grouping approach. Massive datasets are no problem for this approach. The K-means clustering outcome is influenced by the initial points in the optimization process. Cluster centres should be chosen at random for each cluster. These centres should be as widely apart as feasible. The clustering process and outcomes are influenced by the starting points used. Effective cluster assignment is made possible in large part by the Centroid initialization. As a result of the initial centroid values assigned, the clustering convergence behavior is also reliant on these values. In order to improve the clustering performance of the K-Means clustering method, this work focuses on the assignment of cluster centroid selection. An initial cluster centroid is assigned by using centres derived from partitioning of data along the data axis with the highest variance, as described in this study. The experimental findings show that the suggested method is superior to the standard method in terms of clustering outcomes.