Abstract

Ability of ideal bone mineral density (BMD) in athletes especially within the female athletes is an worthy attribute of confirming their good health (safety) through their lives, and examining of BMD is necessary and also a prime objective to circumvent or evade fractures and bone-related injuries. Several tools for assessing bone health exist today. However, lacking the applicability needed for constant watching and also not capable for the female athletes’ demographics. As a consequence of this, we explore the main goal of utilizing the machine learning`s supervised binary classifier techniques to discriminate concerning standard and low-bone-mass individuals amongst female athletes (games and sports) using feature manifestations extricated as of the given feedback form (usually questionnaire). Dataset comprised >200athletes. We evaluated five distinct models: decision-tree (DT), logistic-regression (LR), multi-layer perceptron (MLP) random-forest (RF), and XG-Boost. The data validation done through cross-validation plus significancy of the features were measured via the imperative permutation. XG-Boost showed the most balanced results in terms of sensitivity and specificity, achieving values of 0.94 and 0.63 which was also obtained an area under curve AUC) of 0.74 and an accuracy of 0.68. We examined that the duration of the current period of amenorrhea, and impact of sport, showed the highest relevance, and was coherent with preceding literature. Other features such as thinness level, number of training days in a week and age at menarche also showed high importance. The models demonstrated promising results in identifying low bone mass subjects from normal ones, indicating that the feature-manifestations based on questionnaires can be an important source for evaluating low BMD in female athletes.