Predictive Performance of Machine Learning on Low-Birth-Weight Classification: A Study from Asia Developing Countries

This study is aimed to evaluate the predictive performance of several machine learning models in classifying low birth weight (LBW) infants. This classification is necessary, as low birth weight is linked to many health hazards for newborns. This study conducted machine learning to examine socio-economic variables, maternal health, and additional pertinent aspects that influence low birth weight (LBW) in developing countries, such as India, Indonesia, Jordan, and the Philippines. The independent variables were type of residence, number of household members, mother's education level, mother's occupation, father's occupation, welfare status, number of births for the last 5 years, mother's age at first birth, mother's smoking status, birth order, infant's alive status, number of antenatal care visit, and type of antenatal care. The total eligible sample included 12,393 respondents of Indonesia, 21,681 of India, 6,365 of Jordan, and 5,704 of the Philippines. The findings demonstrate that several machine learning models, including Support Vector Machines (SVM), Random Forest, and Decision Trees, exhibit differing degrees of accuracy in predicting low birth weight (LBW) across India, Indonesia, Jordan, and the Philippines. For example, SVMs exhibited superior performance, although Naive Bayes attained elevated sensitivity. The results indicate that customized strategies reflecting regional attributes are necessary for enhancing prediction precision in LBW classification. This underscores the need of accounting for local socio-demographic variables when using machine learning models in healthcare study.