["Australasian Journal on Ageing, Volume 45, Issue 2, June 2026. ", "\nABSTRACT\n\nObjective\nAccurate estimation of skeletal muscle mass is a key component in the screening for sarcopenia. Anthropometric parameters provide a non‐invasive, cost‐effective alternative, yet their predictive capability can be enhanced with machine learning. This study aimed to develop and evaluate machine learning models to predict skeletal muscle mass measured by bioelectrical impedance analysis (BIA) in individuals aged 70 years and older using anthropometric measurements and demographic data.\n\n\nMethods\nAnthropometric data from 984 older adults aged 70 years and older were obtained from the 8th Size Korea Human Body Measurement Survey. Data preprocessing included removing missing values, applying Min‐Max scaling and selecting relevant features. Machine learning models were developed, optimised using a tuning framework and evaluated through error metrics, residual analysis and feature importance interpretation.\n\n\nResults\nAmong the machine learning models, XGBoost achieved the highest predictive accuracy (RMSE = 1.29, R2 = 0.92) while using the fewest predictors (18), demonstrating efficiency and effectiveness. Key features included calf circumference, gender, weight, height and waist circumference. SHapley Additive exPlanations analysis emphasised the dominant role of calf circumference, reinforcing its predictive capability for skeletal muscle mass.\n\n\nConclusions\nThis study highlights the potential of machine learning for predicting skeletal muscle mass using non‐invasive anthropometric measurements. These findings support the proposed model as a scalable and cost‐effective tool for estimating skeletal muscle mass, which is a key quantitative component of sarcopenia screening.\n\n"]