Maternal health is a critical public health concern globally, with maternal mortality and morbidity rates posing significant challenges. This project aims to explore maternal health risk data and leverage machine learning models to address this issue effectively. We delve into techniques to understand and predict maternal health risks, utilizing data science tools and methodologies.
The dataset used in this project, "Maternal Health Risk Data," sourced from Kaggle, encompasses vital attributes such as age, blood pressure (systolic and diastolic), blood sugar levels, body temperature, heart rate, and pregnancy risk levels. Preprocessing steps, including handling missing values, standardizing features, and label encoding, were undertaken to ensure data quality and suitability for analysis.
Visualizations such as box plots, histograms, and scatter plots were employed to explore relationships between variables and gain insights into the dataset's characteristics. These visualizations aided in understanding the distribution of features and identifying potential patterns and correlations.
Various machine learning algorithms, including Random Forest Classifier, Logistic Regression, and Linear Regression, were explored to predict pregnancy risk levels. Evaluation metrics such as accuracy scores and mean squared error (MSE) were utilized to assess model performance.
- Random Forest Classifier: Accuracy Score ≈ 0.84
- Logistic Regression: Accuracy Score ≈ 0.625
- Linear Regression: Accuracy Score ≈ 0.648
Random Forest Classifier and Linear Regression emerged as the top-performing models, exhibiting promising results in predicting pregnancy risk levels.
The analysis revealed the significance of factors such as age, blood pressure, body temperature, and heart rate in predicting pregnancy risk levels. Although achieving promising accuracies, further optimization of models is essential to enhance predictive performance, particularly concerning blood sugar level prediction. Collaborative efforts across academic disciplines are imperative to address maternal health disparities globally and refine predictive models for improved maternal healthcare outcomes.
This project underscores the importance of leveraging data science and machine learning techniques to mitigate maternal health risks effectively. By incorporating additional variables and refining predictive models, we can enhance their utility in maternal healthcare decision-making. Future research endeavors should focus on comprehensive data integration and collaborative initiatives to drive innovation and improve maternal health outcomes worldwide.
-
Ahmed, M., Kashem, M. A., Rahman, M., & Khatun, S. (2020). Review and Analysis of Risk Factor of Maternal Health in Remote Area Using the Internet of Things (IoT).
-
Amore, A. D., Britt, A., Arconada Alvarez, S. J., & Greenleaf, M. N. (2023). A Web-Based Intervention to Address Risk Factors for Maternal Morbidity and Mortality (MAMA LOVE).
-
Centers for Disease Control and Prevention. (2019). Racial/ethnic disparities in pregnancy-related deaths - United States, 2007–2016.
-
Kaggle: Maternal Health Risk Data.
-
World Health Organization. (n.d.). Maternal Mortality.
- Data Mining with Python