Predictive Modeling of Body Fat Percentage Using Multivariate Anthropometric Measurements and Density Assessment

An accurate Body Fat Percentage (BFP) assessment is required to evaluate an individual's health status and risk for several diseases. Traditional plans for BFP determination, namely dual-energy X-ray absorptiometry (DXA) and hydrostatic weighing are highly authentic, though often impractical due to their cost and complexity. This research focused on building a predictive model for BFP using readily available anthropometric measurements combined with density estimation.Our model used standard data consist of 252 samples, that included measurements such as weight, height, waist circumference, hip circumference, and skinfold thickness, alongside density assessments derived from air displacement plethysmography (ADP). And with feature engineering, important features are selected, like body mass index, hip ratio, and log values, and trained an optimal ANN model, after training a complete result analysis and interpretation is provided. This model indicate that the integration of multivariate anthropometric measurements with density assessment especially increase the accuracy of BFP predictions. The best-performing model acquired a MAE of 1.5% and an R² of 0.92, exceeding traditional anthropometric equations.