Free-parameters clustering of spatial data with non-uniform density
Dongquan Liu, Olga Sourina · 2005
Clustering is a challenging task due to the rapid increase of data variety and the lack of prior knowledge about data. On the other hand very few clustering methods can successfully and automatically deal with the irregular data sets where density varies not only across clusters but also inside clusters, and clusters can also be linked by multiple bridges. Thus, it is important to design a clustering method that can handle such irregular data sets and generate all values of parameters automatically. In this paper, we proposed a new Automatic Nonuniform Density Clustering algorithm (ANDC) based on triangulation method. This approach allows us to cluster irregular data sets efficiently finding uniform and nonuniform density clusters as well. The method does not require any input from the user. The results of tests and comparisons with other algorithms shown in this paper confirm the efficiency of our method.