AI-Based Maternal Health Risk Prediction System Using Machine Learning

Abstract

Maternal mortality continues to be a public health challenge in developing countries, especially in rural and semi-urban areas where regular prenatal monitoring is not available. This paper presents an Artificial Intelligence (AI) based system for prediction of maternal health risk using machine learning techniques which classifies pregnancy risk levels as Low, Moderate or High. The system uses a real-world clinical dataset of 6103 patient records with 8 important physiological parameters: age, body temperature, heart rate, systolic and diastolic blood pressure, BMI and blood glucose levels. The data was trained and evaluated on Random Forest, XGBoost, Gradient Boost and a Voting Ensemble classifier. The XGBoost model performed the best with 99.26% classification accuracy, and 0.99 precision, recall and F1-scores for all three risk categories The system is implemented as a full-stack web application, using React.js for the frontend, FastAPI for the backend, and Supabase as a cloud-based PostgreSQL database, providing real-time risk assessment and accessible healthcare monitoring for pregnant women.

Introduction

Maternal health focuses on the well-being of women during pregnancy, childbirth, and the postpartum period. Maternal mortality remains a major challenge, particularly in low- and middle-income countries such as India, where limited access to healthcare services, especially in rural areas, increases the risk of pregnancy-related complications. Early identification of conditions such as gestational diabetes, hypertension, and preeclampsia is essential for reducing maternal and infant deaths.

To address this issue, the paper proposes an AI-Based Maternal Health Risk Prediction System that enables pregnant women to enter daily health parameters through a web application and receive instant risk assessments. The system classifies patients into Low, Moderate, or High Risk categories and provides appropriate healthcare recommendations.

Literature Review Findings

Previous studies have explored:

• Machine learning models for predicting gestational diabetes.
• Digital health platforms for remote monitoring.
• Blood glucose prediction using AI algorithms.
• IoT-based maternal health monitoring systems.
• Decision-support systems for gestational diabetes management.

While these studies demonstrated the usefulness of AI and digital health technologies, many lacked continuous monitoring, pregnancy-specific risk assessment, real-world validation, or cost-effective implementation.

Proposed Methodology

• Dataset: 6,103 maternal health records from a clinical dataset containing eight physiological parameters: age, body temperature, heart rate, systolic BP, diastolic BP, BMI, HbA1c, and fasting blood glucose.
• Preprocessing: Label encoding of risk categories and an 80:20 train-test split using stratified sampling.
• Models Evaluated:
  • Random Forest
  • XGBoost
  • Gradient Boosting
  • Voting Ensemble
• System Architecture:
  • Frontend: React.js + Vite
  • Backend: FastAPI
  • ML Layer: XGBoost model
  • Database: Supabase
  • Deployment: Vercel (frontend) and Render (backend)

Results

The XGBoost model was selected for deployment due to its high accuracy and efficiency. It achieved:

• Precision: 99–100%
• Recall: 99–100%
• F1-Score: 99–100%
• Cross-validation accuracy: 99.23%

Key Findings

• The most influential predictors of maternal risk were:
  • Fasting Blood Glucose (31.2%)
  • Systolic Blood Pressure (19.8%)
  • HbA1c (18.7%)
• The system successfully identified both low-risk and high-risk pregnancies using real patient records.
• It provides a scalable, web-based, device-independent solution for early maternal risk detection and healthcare decision support.

Conclusion

We present an AI-based maternal health risk prediction system that integrates a high accuracy classifier (XGBoost) with a fully deployed full-stack web application. The system was trained on a real clinical dataset of 6,103 patient records and reached 99.26% classification accuracy with good precision, recall and F1-scores for all risk categories. The deployed system provides real-time accessible risk assessment for pregnant women and is particularly relevant for underserved rural healthcare settings. Future work will involve the integration of IoT-based sensors for automated vital sign acquisition, expanding the dataset through collaboration with local healthcare institutions, developing a dedicated mobile application and including a healthcare provider dashboard for remote patient monitoring. We are also working on multilingual support to improve accessibility for non-English speaking users in rural communities.

References

[1] S. N. Khorshid, M. J. Reddy, and P. Gupta, “Machine Learning Based Prediction of Gestational Diabetes Mellitus,”International Journal of Medical Informatics, vol. 112, pp. 1– 8, 2018. [2] Kumar and R. Patel, “Digital Health Monitoring System for Diabetes Management,” International Journal of Engineering and Technology, vol. 7, no. 3, pp. 245–250, 2019. [3] J. Brown, L. Smith, and K. Taylor, “Application of Machine Learning Algorithms for Blood Glucose Level Prediction,”IEEE Access, vol. 8, pp. 123456–123465, 2020. [4] P. Ramesh and S. Meena, “IoT and Machine Learning Based Health Monitoring System for Pregnant Women,”International Journal of Advanced Research in Electronics and Communication Engineering, vol. 9, no. 5, pp. 412–417, 2021. [5] M. Alsharif and N. Alotaibi, “Digital Health and Machine Learning Framework for Maternal Health Monitoring,”IEEE Journal of Biomedical and Health Informatics, vol. 26, no. 4, pp. 1765–1774, 2022.

Copyright

Copyright © 2026 Namitha M, Niriksha G Y, R Kanika, Sonu N, Minugu B. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.