Identifying Body Size Group Clusters from Anthropometric Body Composition Indicators

Abstract

Mining of anthropometric data by clustering technique would categorically classify homogenous body size group. The objective of this study is to classify homogenous human body size according to anthropometric body composition indicators. Anthropometric data was measured from 382 men and 391 women of Orissa, India. Percent body fat was calculated from the skinfold parameters. Cluster analysis was then applied with self reported age, stature, weight and percent body fat. The clusters formed were tested statistically for their validity of formation. Multivariate analysis considering age, stature, weight and percent body fat revealed significant differences among men and women (p<0.001). Expectation Maximization (EM) estimated five clusters for men and women by age, stature, weight and percent body fat. Consequently, k-means cluster algorithm was used and it formed five clusters with cumulative increment in stature, weight and percent body fat, for both men and women. However, age individually, did not influence the body size indicators. The clusters were named small, medium, large, X-large and XX-large. Silhouette plot validation of clusters reveals that for both men and women, 95.5% and 98.7% of data, respectively were well-clustered. These cluster results further can generate classification rules to categorize subsequent unseen cases, and may aid in anthropometric database creation, nutritional status, body growth research, etc.