Cluster size prediction for military clothing using 3D body scan data

Aim: To determine how anthropometric characteristics cluster in the New Zealand Defence Force, and to describe the characteristics of each cluster. This information can inform the development of new uniform sizing systems for the New Zealand Defence Force. Methods: Anthropometric data (n = 84 variables) from 1,003 participants (212 females; 791 males) in the New Zealand Defence Force Anthropometry Survey (NZDFAS) were used. The dataset was stratified by gender and variables isolated based on their relevance to shirt and trouser sizing. Principal Component Analysis was used to identify the most important variables for clustering. A combination of two-step and k-means clustering was used to derive cluster characteristics. Results: The PCA identified optimal clothing (shirt = body height and waist girth; and trouser = inseam length and hip girth for females; inseam length and waist girth for males) variables. Two-step and k-means clustering identified optimal cluster numbers of 6 and 10 for female and male clothing, respectively. The female clothing clusters were more variable (intra-cluster) and further apart (inter-cluster) compared to males. Conclusions: Anthropometric measurements in combination with clustering techniques show promise for partitioning individuals into distinct groups. The anthropometry dimensions associated with each cluster can be used by the garment industry to develop specific sizing systems for the New Zealand Defence Force population.