AN AVERAGE-BASED APPROACH FOR INITIAL CENTROID SELECTION IN KMEANS ALGORITHM

Abstract

The underlying research work was focused on one of the standard k-means issue of
initial centroid selection. An average based approach was used for avoiding random cluster
initialization. The experiments of this study showed that the results obtained with proposed method
were better and consistent. It was concluded that the proposed method had less classification error,
reduced total number of iterations and took less execution time than random initialization method.