A human-centered framework for data-driven anthropometric sizing: design and validation in a military context

Traditional anthropometric methods for personalising equipment in high-stakes professions are often costly, time-consuming, and lack scalability. This study proposes and validates a low-cost, human-centered framework that integrates machine learning with usability evaluation to address this problem. The framework consisted of two stages: first, applying clustering algorithms to anthropometric data to establish a data-driven sizing model; and second, developing a smartphone-based prototype to validate the framework's real-world applicability. A comprehensive evaluation with 20 university students and a supplementary validation with 6 active Air Force Academy students demonstrated the framework's success, achieving an average System Usability Scale (SUS) score of 83 (82.5) and a total Questionnaire for User Interaction Satisfaction (QUIS) score of 209.55 (187.33). The data model was also validated, with key anthropometric variables effectively stratifying complex body types (p < .001). The primary contribution of this study is a generalisable framework for developing user-accepted personalised fitting systems in resource-constrained settings.