Predicting Pregnancy Delivery Outcomes Using Machine Learning Algorithms
Loading...
Date
2025-04
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Kenyatta University
Abstract
Pregnancy and delivery outcomes, including live births, induced abortions, miscarriages, and stillbirths, posed a serious threat to world health, especially in areas with poor access to high-quality medical treatment. By creating a machine learning model that precisely forecast these pregnancy delivery outcomes, this study sought to improve mother health and lower perinatal deaths in Kenya. One major problem was that many expectant mothers did not follow advised antenatal care (ANC) procedures, which increased the risk of unfavorable outcomes. By examining the demographic variables that affect adherence to advised ANC procedures and creating a prediction model to identify women at higher risk, this study tackled this issue. The study used data from the 2022 Kenya Demographic and Health Survey (KDHS), a comprehensive dataset. Several machine learning techniques were assessed to find the best method for this prediction task. These methods included Linear Regression (LR), Support Vector Machines, Artificial Neural Networks (ANN), and stacking ensemble learning. A variety of criteria were used to evaluate each model's performance, including sensitivity, specificity, and accuracy. Additionally, the main demographic characteristics linked to the outcomes of pregnancy and delivery among Kenyan women were determined using odds ratios and p-values at a 5% significance level. The Random Forest Gini index was used to determine the relative importance of each predictive variable. With a prediction accuracy of 87.2%, the stacking ensemble model outperformed ANN in pregnant delivery prediction, with ANN coming in second at 83.7%. According to the findings, several factors were significant predictors of pregnant delivery outcomes in Kenya. These included the mother's age, spouse support, education level, wealth index, number of children she had overall, and site of delivery. The results of the study indicated that specific interventions for vulnerable groups, such as low-income and uneducated women, as well as the participation of male partners in prenatal care, may greatly enhance maternal health outcomes. According to the study, pregnant women with low levels of education would benefit from early prenatal care and health promotion initiatives that receive positive media attention.
Description
A Research Project Submitted in Fulfilment of the Requirement of Award of the Degree of Master of Sciences in Computer Science in School of School of Pure and Applied Sciences of Kenyatta University, April 2025
Supervisor;
1.Abraham Matheka