Utility-based data mining: An anthropometric case study

Abstract

One of the most important challenges for the apparel industry is to produce garments that fit the population properly. In order to achieve this objective, it is crucial to understand the typical profile of consumer's bodies. In this work, we aim to identify the typical consumer from the virtual tailor's perspective. To this end, we perform clustering analysis on anthropometric and 3-D data to group the population into clothing sizes. Next, we perform multi-view relational classification to analyze the interplay of different body measurements within each size. In this study, we analyze three different populations as contained in the CAESAR(TM) database, namely, the American, the Italian and the Dutch populations. Throughout this study, we follow a utility-based data mining approach. The goal of utility-base data mining is to consider all utility aspects of the mining process and to thus maximize the utility of the entire process. In order to address this issue, we engage in dimension reduction techniques to find a smaller set of body measurement that reduces the cost and improves the performance of the mining process. We also apply objective interestingness measures in our analysis of demographic data, to improve the quality of the results and reduce the time and search space of the mining process. The analysis of demographic data allows us to better understand the demographic nature of potential customers, in order to target subgroups of potential customers better.