Textile and Apparel Science

Comparing Rectangle Body Shape Using Unsupervised Machine Learning Algorithms

---

Abstract

Key body dimensions to classify female body size and shape can be different from different sample populations. Considering body shape can contribute to better accuracy and fit performance in pattern making and developing sizing systems to cater to a large and diverse population. Therefore, this research explored identifying potential key body dimensions other than the bust, waist, and hip girths and grouping clusters for a rectangle body shape population. Exploratory factor analysis (EFA) and Finite Mixture Model (FMM) with the Expectation-Maximization (EM) algorithm were used to identify interaction patterns among 39 body measurements and categorize rectangle-shaped female subjects into 5 groups based on the frequency of horizontal and vertical measurements and the shoulder slant slope. The shoulder slope could be a potential key dimension to analyze the anthropometric data and develop a sizing system that enables the production of correctly sized clothing for the rectangle-shaped population.

Keywords: Fit and Sizing, Body shape, Key Body Dimensions, Female Figure Identification Technique (FFIT), Unsupervised Machine Learning, Finite Mixture Model (FMM)

How to Cite:

Jung, U., Rothenberg, L. & Istook, C., (2022) “Comparing Rectangle Body Shape Using Unsupervised Machine Learning Algorithms”, International Textile and Apparel Association Annual Conference Proceedings 78(1). doi: https://doi.org/10.31274/itaa.13814