Using Bagging Neural Network to Predict the Factors Affecting Neonatal Mortality

Abstract

Background: The rate of neonatal mortality is one of the main indices of health, treatment, and development in societies. It reflects the quality of nutrition and life of mothers as well as the rate of healthcare services that mothers and children are provided with by societies. This study aimed to identify the factors affecting neonatal mortality by using a bagging neural network in Rapidminer Software. Methods: The study was conducted on 8053 births (including 1605 death cases and 6448 control cases) all over Iran in 2015. Factors such as maternal risk factors, mother’s age, gestational age, child gender, birth weight, birth order, and congenital anomalies were utilized as the predictor variables of the bagging neural network. Some criteria, including the area under the ROC curve, as well as the property and sensitivity of the bagging neural network, were compared with the neural network model. The bagging neural network with 99.24% precision rate enjoyed better results in predicting the factors affecting neonatal mortality. Results: Our suggested method revealed that gestational age is the most significant predictor factor of a neonate's status at birth time. Besides, 1-minute Apgar, need for resuscitation, 5-minute Apgar, birth weight, congenital anomalies, and birth order, as well as diabetes and preeclampsia in mothers were identified as the most significant predicting factors after the gestational age. Conclusion: Factors discovered in this study can be considered to decrease neonatal mortality. This can help the health of mothers’ community, optimize healthcare services, and development of societies.