The ability of machine learning models to predict the exclusive breastfeeding

Background Identifying factors that influence breastfeeding is essential because of its importance for the health of both mothers and infants. Our study intended to apply machine learning (ML) models to assess the clinical, demographic, and obstetric factors that predict exclusive breastfeeding (EBF). Methods Through a prospective cohort study of infants born in Bandar Abbas, Iran from January 1, 2022, to December 31, 2022, we created ML models that predict the EBF within six months. Our dataset was obtained from the “Iranian Maternal and Neonatal Network. Infants with gestational age under 34 weeks, absent data on gestational age or birth weight, and the death or relocation of the infant or mother within the initial months after birth were excluded. The outcome measure was EBF within six. factors that may be associated with EBF were preliminaries identified to run the ML models. We utilized several metrics to assess the performance of the models such as the area under the curve (AUC), accuracy, precision, Brier score, recall, F1 score, precision-recall (PRAUC). Results The prevalence of EBF in our study was 60.9%. The most influential predictors across all ML models were maternal age, maternal education and occupation, gestational age, parity, mode of delivery, newborn Apgar score, newborn weight, presence of a doula, skin-to-skin contact, participation in birth classes, prior breastfeeding experience, home assistance, NICU admission, chronic maternal disease, BMI, and multiple pregnancies. Among all ML models XGBoost stands out as the top model across multiple measures with an accuracy of 0.6, AUC of 0.59, a Brier Score of 0.2, and a PR AUC of 0.7, reflecting the least error and highlighting this model’s superior accuracy and dependability in predicting EBF. Previous breastfeeding experience significantly influences the model's predictions. It suggests that a mother's past breastfeeding experience is essential in influencing her choice to keep breastfeeding in her current childbirth. Maternal job, participation in birth classes, type of delivery, and gestational age are among the strongest predictors of EBF. Conclusions A strong model was created to predict EBF, highlighting previous breastfeeding experience as a key predictor. XGBoost has shown acceptable performance in predicting EBF.