The accuracy of non-invasive prenatal testing (NIPT) is highly dependent on the concentration of fetal cell-free DNA in maternal plasma, with Y chromosome concentration serving as a key indicator for assessing male fetus samples. To address the clinical risks arising from individual differences among pregnant women overlooked by current standardized protocols, this study developed a data-driven framework for optimizing personalized testing timing. First, a hybrid GAM-RF predictive model integrating a generalized additive model (GAM) with random forest (RF) was constructed to accurately capture the complex nonlinear relationships among multiple factors, including maternal BMI, gestational age, maternal age, and Y chromosome concentration. The model demonstrated superior performance on the test set, achieving a coefficient of determination (R²) of 0.5410, significantly outperforming traditional models. Second, data-driven heterogeneous subgrouping of pregnant women was performed using a model-based decision tree (MOB) algorithm. Finally, a multi-objective programming model was established with the goal of minimizing clinical risk, thereby determining the optimal testing time points for different BMI subgroups: 11.0 weeks for the low-BMI risk group, 14.1 weeks for the medium-BMI risk group, and 21.1 weeks for the high-BMI risk group. The methodology proposed in this study bridges the gap from “prediction” to “decision-making,” providing a methodological framework and decision-making basis for the personalized and precise implementation of NIPT. Furthermore, this analytical framework-encompassing accurate biomarker prediction, population stratification, and multi-objective optimization-offers guidance for the development and optimized calibration of next-generation smart textiles and wearable biosensors to enable personalized health monitoring.

y chromosome concentration, GAM-RF hybrid model, personalized testing timing optimization, intelligent textiles