Sizing for the clothing industry using principal component analysis - Australian example

Abstract

<jats:title>Abstract</jats:title> <jats:p>Principal component analysis was used to condense twelve anthropometric measurements into the first two principal components, which maximize information relevant to setting up a size and shape specification for the development of well-fitting apparel patterns, in this case for the manufacture of a bodice. The use of only two principal components allows for the specification of a practical sizing system. The database for the work consists of 54 measurements taken from 1330 adult women ranging in age from 18 to 70 + years measured in six capital cities in Australia in 2002 and 2004. The twelve measure subset was chosen by an experienced apparel pattern maker to be most relevant for a construction of a bodice pattern. Additionally, the measures were ordered in importance using all pair wise comparisons. This aspect of the work is essential in maximizing the retention of information using only two numbers. A new feature uses the scale dependency of principal component analysis to influence the accuracy of the reconstructed data using only the first two principal components. The paper defines 36 categories: twelve sizes and three shapes in each size. Percentages of women in the 36-size shape groups are determined. In as far as the sample of 1330 women is representative of the Australian population, the percentages provide valuable information for the apparel industry. The main purpose of the paper is to provide an example of how it is possible to improve sizing for clothing based on anthropometric body measurements in a manner acceptable to the apparel industry. The paper shows the body scans of a subset of nine actual women who illustrate the size and shape variations existing in this sample.</jats:p>