GROUPABIL1TY: an analysis of the properties of binary data matrices for group technology

Abstract

Block-diagonalization of the machine-component incidence matrix is the first step in the implementation of group technology. Even powerful algorithms will fail to achieve this if the matrix itself is not amenable to block-diagonalization. The present work analyses the properties of the matrix and identifies the standard deviation of the pairwise similarities (Jaccard7rpar; of the vectors as the major factor that decides the groupability of the data set. Many data sets ranging from the perfectly groupable to the most ill structured ones are analysed and presented. The groupability curves show the variation of the property against the relevant factors.