A Median Based External Initial Centroid Selection Method for K-Means Clustering

Clustering is one of most prevalent data mining techniques that groups objects of same type. The applications of clustering are enormous in fields like medical, business and education especially in medical diagnostics such as grouping of patients at same stage, having common diseases and so on. K-means is one of the best known and widely implemented clustering algorithm that has been greatly suffering from the selection of random initial centroids which degrades the performance of the clustering results. Thus, in this work novelmedian based initial centroids have been generated and imposed onto an experimental dataset to analyze the performance of the proposed work. The results have shown that the proposed work, improved the accuracy of clustering with reduced number of iterations.