Skip to Main Content
In this paper, a dynamic process for data clustering is presented. It is based on the collective behavior among the objects of the input dataset. Each object is assigned an energy state, so they interact with each other by exchanging their energy, causing similar objects to take similar states. Finally, a classical algorithm such as k-means is applied on the energy vectors to actually cluster the data. Experiments show that the energy exchanging process is able to transform complex arrangements of objects into arrangements much easier to cluster. Moreover, the energy exchanging process is resilient to the mixture of clusters to some extent.