MACHINE LEARNING PREDICTION MODELS OF BIRTH WEIGHT OF NEW BORN BABIES IN FCT ABUJA, NIGERIA

Authors

  • Alegbe A.A. Department of Statistics, Nasarawa State University, Keffi,
  • Adenomon M.O. Department of Statistics, Nasarawa State University, Keffi,
  • Maijamaa B. Department of Statistics, Nasarawa State University, Keffi,

Abstract

This research aimed at creating a machine learning model for predicting birth weight using the maternal risk factors that have been found to be associated with low birth weight. The data covered a period of ten years from 2010 to 2019 was utilized where the variables were extracted from the births recorded file. The study population included mothers between the age of 15 to 49 years. The machine learning algorithms employed were logistic regression, Decision trees, Random Forest, Support Vector Machines, Gradient Boosting and K -Nearest Neighbors, Neural Network, Gradient boosting and Linear Regression. The metrics used for classification method were Accuracy, Sensitivity, Specificity and Kappa. In terms of accuracy, the best machine learning model was the Decision tree with an accuracy of 0.9823. The other five models produced an accuracy that ranged between  0.9806 to 0.9822. Based on the kappa, decision tree again emerged to be the best with a value of 09631. The rest of the models had a kappa that ranged from 0.8859 to 0.9592. Sensitivity was also evaluated and Neural Network and support vector machine had the same sensitivity value of 0.9941 whereas the other models managed a recall score ranging from 0.9501 to 0.9853. Moreover, Specificity was also examined. Logistic Regression model had the best specificity value of 0.9908. The rest of the models ranged from 0.9378 to 0.9778. Furthermore, the ROC curves of all the tested models were plotted and the area under the curved evaluated. The decision tree had the highest area under the curve of 0.9896. The AUC of the other models ranged from 0.9440 to 0.9816. Therefore, from these results based on the performance metrics and ROC-AUC, decision tree emerged to be the most robust model for classification method.  Furthermore, the ROC-AUC was used to test the classification ability of the models to differentiate between the low-birth-weight cases and the cases without low birth weight.

Downloads

Published

2024-12-30

Issue

Section

ARTICLES