Gestational diabetes mellitus (GDM) is a major disease affecting pregnant women. Screening for GDM and applying adequate interventions may reduce the risk of adverse outcomes. The diagnosis of GDM depends on OGTT tests performed in the late second trimester. The goal of this study was to create a hybrid model for the prediction of GDM in early pregnancy in women using a machine learning algorithm using polling-based feature selection techniques. Methods: Data on 1725 pregnant women in early gestation were used to fit the GDM risk-prediction model. Predictive maternal factors were selected through the poling method of the feature selection model. Predictive maternal factors were selected through the poling method of feature selection. Incorporated selected maternal factors into a modified Naive Bayes and decision tree. The area under the receiver operating characteristic curve (AUC) was used to assess discrimination. Results: The risk of GDM could be predicted with OGTT zero min, C-peptide HOMA, maternal age, prepregnancy body mass index (BMI), parity, and offspring birth weight with a predictive accuracy of 87.92% and an AUC of 0.766 (95% CI 0.731, 0.801). Conclusions: This GDM prediction model is potentially applicable to alternative decision support systems and women who plan to conceive a baby.

#