Comparative Analysis of Prognostic Model for Risk Classification of Neonatal Jaundice using Machine Learning Algorithms
Peter Adebayo Idowu, Ngozi Chidozie Egejuru, Jeremiah Ademola Balogun, Olusegun Ajibola Sarumi
This study focused on the development of a prediction model using identified classification factors in order to classify the risk of jaundice in selected neonates. Historical dataset on the distribution of the classification of risk of jaundice among neonates was collected using questionnaires following the identification of associated classification factors of risk of jaundice from medical practitioners. The dataset containing information about the classification factors identified and collected from the neonates were used to formulate predictive model for the classification of risk of jaundice using 2 machine learning algorithm – Naïve Bayes’ classifier and the multi-layer perceptron. The predictive model development using the decision trees algorithm was formulated and simulated using the WEKA software. The predictive model developed using the multi-layer perceptron and Naïve Bayes’ classifier algorithms were compared in order to determine the algorithm with the best performance. The result shows that 10 variables were identified by the medical expert to be necessary in predicting jaundice in neonates for which a dataset containing information of 23 neonates alongside their respective jaundice diagnosis (Low, Moderate and High) was also provided with 22 attributes following the identification of the required variables. The 10-fold cross validation method was used to train the predictive model developed using the machine learning algorithms and the performance of the models evaluated. The multi-layer perceptron algorithm proved to be an effective algorithm for predicting the diagnosis of jaundice in Nigerian neonates with a value of 100%.
Share this article