A novel rough value set categorical clustering technique for supplier base management

Abstract

Significant business implications and effective handling of supply side exceptions require a successful Supplier Base Management (SBM). The process of clustering manages the number of suppliers by grouping them on the basis of similar characteristics that reduces the number of variables impacting the operations. Several existing categorical clustering techniques for such grouping contributed well than their predecessors however, the accuracy, uncertainty, entropy and computation are key measures need to be improved. Especially, the existing clustering techniques cluster randomly in case of independent and insignificant type of data. The aim of this research is to introduce a novel rough value set based categorical clustering technique named Maximum Value Attribute (MVA). The proposed MVA techniques overcome the issues of existing techniques by combining the concept of Number of Automated Clusters (NoACs) with rough value set which makes it novel and significant clustering idea. Few relevant and necessary propositions are illustrated to prove the effectiveness of NoACs. The existing and proposed rough sets based and classical categorical clustering techniques are compared in terms of purity, entropy, accuracy, rough accuracy, time and iterations. Experimental results based on a SBM and fifteen (15) benchmark data sets reveal better performance of MVA. The experimental results show significant overall percentage improvement of proposed MVA technique against existing rough based techniques for iterations (99.7%), time (99.4%), number of obtained clusters (84%), purity (32%), entropy (32%), accuracy (20%), and rough accuracy (13%).