Predicting Low Birth Weight in India Using Machine Learning: Insights from NFHS-5

VIKAS KAMBLE, International Institute for Population Sciences (IIPS)
Murali Dhar, International Institute for Population Sciences (IIPS)

Low birth weight (LBW) reduces a new-Born's productivity as an adult and increases their chances of morbidity and Infant death. The World Summit for Children has chosen the prevalence of LBW as a key metric for tracking crucial health. In developing nations, LBW is one of the leading causes of infant death. This work aims to identify the main cause of LBW and predict LBW children’s. The study sample, which was focused on infants and was derived from NFHS-5 data, was further analyzed using machine learning techniques. The models were trained using 80% of the data and tested on the remaining 20%. The assessment metrics have been used to evaluate and compare the performance predictive models. Since our data were highly imbalanced, we assessed the performance of each classifier. The best performance was achieved using the Support Vector Machine classifier with an accuracy of 90.60% which is higher than other classifiers used in this study. The Logistic Regression, Decision Tree, KNN, Random Forest and Gaussian classifiers have accuracy values of 90.36%, 89.28%, 88.38%, 89.21%, and 89.64%. The mother's age plays a crucial role in determining LBW, as identified using the best SVM model.

Keywords: Bayesian methods , Data visualisation , Population projections, forecasts, and estimations, Multi-level modeling

See extended abstract.