Study of a cluster algorithm based on rough sets theory

Clustering in data mining is a discovery process that groups a set of data so that the intra-cluster similarity is maximized and the inter-cluster similarity is minimized. Existing clustering algorithms, such as k-medoids, are designed to find clusters, but these algorithms will break down if the choice of parameters in the static model is incorrect with respect to the data set being clustered Furthermore, these algorithms may break down when the data consists of clusters that are of diverse shapes or densities. Combined the method of calculating equivalence class in rough sets, an improved clustering algorithm based on k-medoids algorithm was presented in this paper. In this algorithm, the number of clusters was firstly specified and the resulting clusters were returned via the k-medoids algorithm, and then the clusters were merged using rough sets theory. The illustrations show that this algorithm is effective to discover the clusters with arbitrary shape and to set the number of clusters, which is difficult for traditional clustering algorithms.