Comparative Analysis of ML Algorithms for Predictive Prenatal Monitoring
Document Type
Conference Proceeding
Publication Date
2022
Abstract
Despite the rapid advancements in prenatal health care services a lot of low-income sectors are experiencing high fetal mortality rates because of inaccessible prenatal health services. The reasons include financial incapability, pregnant women residing in isolated regions cannot access reliable healthcare services, and insufficient healthcare equipments in certain areas. To assess the shortcomings of the current prevalent methods, this study proposes an accurate non-invasive process of prenatal health care assessment by using a trained machine learning algorithm in a telemedicine setup. This setup uses a mobile app for patients and doctors connected to a cloud storage database where the patient information is stored. The predictive model would then be able to predict whether a patient is a high-risk or low-risk pregnancy based on the patient information inputted in the app. The Machine Learning algorithms to be compared are Random Forest Decision Tree, Decision Tree, $\mathbf{K}$ -nearest neighbor, and SVM. After pre-processing the dataset, the predictive model was created by inputting the dataset of patient information to multiple machine learning algorithms and assessing their performance parameters. Based on the testing results, the preferred algorithm to be used is the Random Decision Tree Algorithm which had better overall performance than the previous model of Bautista and Quiwa. The study showed the further potential of the Machine Learning algorithm as a healthcare tool as data can now be easily attained using current technologies. Further improvement with the telemedicine setup could aid women who do not have sufficient access to healthcare services.
Recommended Citation
Bautista, J.M., & Reyes, R.S.J. (2022). Comparative Analysis of ML algorithms for Predictive Prenatal Monitoring. 2022 IEEE 4th Eurasia Conference on Biomedical Engineering, Healthcare and Sustainability (ECBIOS), 148-151. https://doi.org/10.1109/ECBIOS54627.2022.9945019
