Maternal Health Risk Prediction Using Machine Learning Methods

Abstract

This work employed machine learning methods – decision tree algorithm, support vector machine (SVM) and logistic regression to predict health risk outcome of pregnant women based on their vital signs. The aim was to determine which of the model most suitable for the prediction of health risk for any pregnant woman. Evaluation of the methods were done using Accuracy, Precision, F1-Score and recall. The results revealed that the Decision tree model achieved the highest accuracy of 0.8669 (86.99%), that indicates correct prediction in 86.99% of the cases. Also, it achieved the highest precision for all the risk categories (high_risk, low_risk and mid_risk) with values (87%, 86% and 75% respectively) implying lower likelihood of false positive predictions for all risk categories. The Decision Tree model appears to be a promising approach for predicting the impact of vital signs on health risk of pregnant mothers. It exhibited high precision, high recall (sensitivity) and a balanced F1-score, suggesting accurate predictions with very low rate of false positives.