Development of a Data-Clustering Method Focusing on Simplicity of Cluster Structures and Its Application to Chemoinformatics

Abstract

A cluster validity index (CVI) called "simplicity index" (SI) is newly proposed to enhance the accuracy of data clustering in machine learning. This index is derived to emphasize the importance of simplicity in cluster structures. The characteristics ofSI and its advantages over the known methods in the literature are discussed. SI is applied to classification of nucleotide sequences of nitrogen-fixing genes.