Mean Centroid K-Means Clustering (MCKM) Based on Boundary Region Analysis for Share Market Database

Abstract

In many research areas it’s always found that it is very difficult to cluster the databases which come under the close region of clusters. This work proposes a novel methodology for performing a structured cluster analysis of share market data. In this work clustering is done using the Mean Centroid K-Means Clustering (MCKM) algorithm and complex regions are selected which are closed to two or more clusters and this selected database is again carefully examined by each of the attribute and then finally clustered to produce more accuracy. In the proposed MCKM Clustering algorithm, centroid value is chosen based on the mean value of the data points. Instead of using random centroid values, mean value of the data points are used as the centroid for clustering. Proposed work has used stock returns at different times along with their valuation ratios from the stocks of National Stock Exchange (NSE) for the India's leading stock exchange. Results of the analysis shows that MCKM analysis builds the most compact clusters when compared to K-Means clustering and self organizing maps (SOM) for share reduced features data. The results are measured by using Silhouette index and Davies–Bouldin index.