Determining the number of clusters in a data set