Adepoju Temitope Elizabeth

Data Analyst | Relationship Officer | Turning Data into Insights and Impact

View the Project on GitHub Temitopeadep/data-analyst-portfolio

SmartMum: Maternal Health Risk Prediction Through AI (XGBoost Model)

Model Accuracy AUC Score

Overview

This project was developed for the 3MTT knowledge showcase (July Edition). It’s an AI-powered maternal health risk prediction tool built using BigQuery ML (XGBoost) and integrated with Power BI and Power Apps. It enables early identification of pregnancy-related complications by analyzing clinical and demographic data

Problem Statement

Many pregnant women, especially in underserved areas, miss early detection of health risks due to limited access to timely screening and data insights. There’s an urgent need for a smart, accessible tool that predicts maternal health risks early, supporting safer pregnancies and informed interventions.

Project Goal

To reduce maternal health risks through early prediction and actionable insights for healthcare providers and expectant mothers.

Project Components

Executive Summary

  1. Model: XGBoost via BigQuery ML.
  2. Accuracy: 89.6%
  3. Recall: 87.3%
  4. Precision: 88.1%
  5. AUC Score: 0.95
  6. Dataset Size: 1,000 Records

    Key Features

    • Uses real health indicators to predict Low / Medium / High risk.
    • Power BI dashboard with demographics, clinical variables, and trends.
    • Power Apps form for public access without technical knowledge.

Data Source

Dataset: Maternal Health Risk | Kaggle

smartmum1

smartmum2

PowerApps Interface

powerapp

Findings

  1. Model Performance: The XGBoost model achieved strong metrics — Accuracy: 89.6%, Recall: 87.3%, Precision: 88.1%, and AUC: 0.95, confirming its reliability in predicting maternal health risks.
  2. Risk Distribution: Most cases fall into the Medium Risk category, suggesting a window for preventive care. High-risk cases, though fewer, remain clinically significant.
  3. ANC Visits: Low or zero antenatal care visits strongly correlate with higher maternal health risk levels. Regular ANC visits are protective.
  4. Clinical Indicators: Elevated BMI, Heart Rate, Blood Pressure, and Body Temperature are key predictors of high-risk classifications.
  5. Demographics & Access: Both urban and rural regions exhibit similar risk trends. However, women with limited access to facilities or education face compounded risks.

Recommendation