One of the weaknesses of the P-median formulation of the cell configuration problem in group technology (GT) is that the number of cells required has to be specified in advance. Therefore, the most natural clustering of the input matrix might not be obtained. This paper suggests an approach for overcoming this weakness by using a new measure of similarity coefficient. Unlike most of the existing measures of similarity coefficient, the proposed measure can take both positive and negative values and therefore identifies both similarity and dissimilarity between machines. The proposed approach was tested on several problems taken from the literature. The results of this computational study demonstrate the superiority of the proposed method.