A Classification Model for Severity of Neonatal Jaundice Using Deep Learning

摘要

Neonatal jaundice is a yellowish discoloration of the white part of the eyes and skin in a newborn baby due to high bilirubin levels. An early diagnosis of the severity of neonatal jaundice using machine learning will decrease neonates' likelihood of developing complications. The study elicited knowledge on the variables that are associated with the severity of neonatal jaundice and collected relevant data from a tertiary hospital in south-western Nigeria. The study formulated the predictive model for the severity of neonatal jaundice based on the variables identified using deep learning with multi-layer perceptron (MLP) classifier for varying number of epochs. The results of the study showed that using the deep learning with MLP classifier and 5 epochs had the lowest error rate however with the highest build time and provided a better model compared to the use of the other number of epochs. The study concluded that the using deep learning with MLP classifier and 5 epochs, the development of the classification model for the severity of neonatal jaundice patients receiving treatment was more effective due to its ability to understand the relationship between the attributes and their respective target class labels.