Investigating the Accuracy of Subjective Clustering and Bootstrap Application to Subjective Clustering Using an Empirical Population

Abstract

For a new product to be successful in today's market, engineers need to identify representative customer needs. One approach to identify representative needs from a large number of needs is Subjective Clustering (SC). A set of clusters obtained from SC is a point estimate of clusters generated by a population of customers. Another approach is to apply Bootstrap (BS) to SC. By applying BS to SC, engineers can draw an inference about population primary clusters. This paper compares the accuracy of estimating population primary clusters using SC and Bootstrap applied to SC (BS-SC). The authors recruited participants to perform the clustering experiments and assumed that these participants consist a population. The authors randomly sampled subsets of participants and evaluated how accurately SC and BS-SC identify population primary clusters. When the sample size is small relative to the population, BS-SC estimated population primary clusters more accurately than SC.