An Improved Spectral Clustering Algorithm Based on Neighbour Adaptive Scale

Spectral clustering algorithms have seen an explosive development over the past years and been successfully used in data mining and image segmentation. They can deal with arbitrary distribution dataset and easy to implement. But they are sensitive to the datasets which include clusters with distinctly different densities and the parameters must be selected cautiously. This paper proposes an improved spectral clustering algorithm based on neighbour adaptive scale, who fully considers the local structure of dataset using neighbour adaptive scale, which simplifies the selection of parameters and makes the improved algorithm insensitive to both density and outliers. Experimental results show that, compared with k-means and standard spectral clustering, our algorithm can achieve better clustering effect on artificial datasets and UCI public databases.