A new measure for assessment of clustering based on kernel density estimation

A new clustering accuracy measure is proposed to determine the unknown number of clusters and assess quality a data set given in any dimensional space. Our validity index applies classical nonparametric univariate kernel density estimation method interpoint distances computed between members data. Being based on distances, it free curse dimensionality therefore efficiently computable for high-dimensional situations where study variables can be larger than sample size. The compatible with algorithm every kind distance defined have function. conducted simulation proves its superiority over widely used cluster indices like average silhouette width Dunn index, whereas applicability shown respect Biostatistical Alon large Astrostatistical application time series light curves variable stars.