The use of a similarity coefficient-based method to perform clustering analysis to a large set of data with dissimilar parts

When performing a clustering analysis in large data sets using a coefficient based-clustering technique, there may be parts that are so dissimilar that they do not join any part or part cluster except at very low threshold values. Therefore, at threshold values where workable number of part families are obtained, an unbalanced configuration in terms of the number of parts per family can be created. As a result, many part families will have a low number of parts and few families will be formed by a high number of parts. This issue leads to the definition of the term “remainder cluster” which is used to designate those parts or cluster of parts that at a certain threshold value do not join any part family and do not have sufficient parts to be considered a feasible part family. In this paper, an approach to deal with the “remainder clusters” is proposed. The objective is to form part families, each with a feasible number of parts.