Growing Hierarchical Self-Organizing Map Computation for Clustering in Cellular Manufacturing

In this paper, a novel and potent approach of the use of Growing Hierarchical Self-Organizing Map (GHSOM) for organization and visualization of machine-part cell formation has been proposed for the first time. The results are compared with the ones obtained by the ART1 algorithm on 14 problems taken from the literature. The equivalent code was written in Matlab language and executes on an Intel Pentium 4 computer. It is found that the proposed GHSOM-based algorithm outperforms the existing ART1 algorithms even for largersize problems both in terms of the group technology efficiency (GTE) and the computational time, the performance measures commonly used in literature. It is found that the proposed algorithm resulted increase in GTE of 10% (approximate) from best result of literature in case of 28.57% of the problem data sets and the outputs of cell formation are either superior (in case of 71.42% data sets) or the same as the existing methods (rest data sets). The outputs of the experiments conducted in this research lead us to the conclusion that the GHSOM is more promising than the traditional ART1 algorithm owing to its adaptive architecture and the ability to expose the hierarchical structure of data