Evaluating machine learning models for clothing size prediction using anthropometric measurements from 3D body scanning

Abstract

Abstract An analysis of a dataset comprising 677 participants revealed substantial discrepancies in size categorization. Only 63 individuals (9.15%) maintained consistency across bust, waist, and hip measurements, whereas 614 participants (90.84%) exhibited size variations, and 35.45% were not adequately accommodated by the existing sizing scheme. These findings highlight significant challenges in garment selection, potentially leading to dissatisfaction and increased return rates. This study evaluated the effectiveness of support vector machine (SVM) and principal component analysis-SVM (PCA-SVM) models for clothing size prediction via 3D body scanning data. The traditional SVM model, which focuses on primary measurements, achieves an accuracy of 89.66%, outperforming the PCA-SVM model (68.97%), which incorporates additional dimensions. These results underscore the effectiveness of SVMs in predicting clothing size categories and emphasise the intricate relationship between body morphology and garment fit.