GDILC: a grid-based density-isoline clustering algorithm

Abstract

A novel clustering algorithm, the grid-based density-isoline clustering (GDILC) algorithm is put forward in this paper. The central idea of GDILC is that the density-isoline figure depicts the distribution of data samples very well. We use a grid-based method to calculate the density of each data sample, and find relatively dense regions, which are just clusters. GDILC is capable of eliminating outliers and finding clusters of various shapes. It is an unsupervised clustering algorithm because it requires no human interaction. The high speed and accuracy of the GDILC clustering algorithm is shown in our experiments.