Visual clustering based on chemical recognition system of ants

This paper describes the Visual AntClust clustering algorirthm that relies on a modeling of the chemical recognition system of ants to build a partition of a data set. The algorithm associates each artificial ant with a data object to be classified and represent its chemical signature in a 2D euclidian space. It then applies rules that mimic the behavior of real ants to group into the same nest (or cluster) the artificial ants. Therefore similar ants (or data) also tends to have similar coordinates in the 2D space. Experimental results show that this method can achieve good performances on artificial and real data sets and allows for a good visualization tool.