TY  - JOUR
AU  - Mazumder, Partha Pratim
AU  - Hasan, Raza
AU  - Mahmood, Salman
AU  - Palaniappan, Sellappan
PY  - 2026
DA  - 2026/04/02
TI  - Predictive Analytics for Maternal Mortality in Bangladesh: An Interpretable ML Framework with Ensemble Methods
JO  - ICCK Transactions on Machine Intelligence
T2  - ICCK Transactions on Machine Intelligence
JF  - ICCK Transactions on Machine Intelligence
VL  - 2
IS  - 3
SP  - 127
EP  - 143
DO  - 10.62762/TMI.2026.182317
UR  - https://www.icck.org/article/abs/TMI.2026.182317
KW  - maternal mortality prediction
KW  - interpretable machine learning
KW  - class imbalance techniques
KW  - SMOTE
KW  - ADASYN
KW  - CTGAN
KW  - Tomek Links
KW  - ensemble learning
KW  - Random Forest
KW  - SHAP explainability
KW  - healthcare risk modeling
KW  - public health analytics
AB  - Maternal mortality in Bangladesh remains a critical public health challenge, with recent evidence indicating stagnation in mortality reduction despite expanded facility-based delivery and skilled birth attendance. Accurate identification of high-risk cases is essential to enable targeted intervention and resource allocation. This study develops an interpretable machine learning framework for maternal mortality prediction using the nationally representative Bangladesh Maternal Mortality Survey 2016 (BMMS-2016). A comprehensive data integration and feature engineering pipeline was implemented across demographic, socioeconomic, and maternal healthcare domains. Given the severe class imbalance inherent in maternal death outcomes, multiple resampling strategies---including Tomek Links undersampling, SMOTE, ADASYN, and CTGAN---were systematically evaluated in conjunction with diverse classifiers and ensemble methods. Among all configurations, Random Forest combined with SMOTE achieved the best overall performance (ROC-AUC: 0.9635; Accuracy: 0.9016; F1-score: 0.8928), demonstrating superior precision--recall balance suitable for rare-event prediction. Tree-based ensemble models consistently outperformed baseline classifiers. Model interpretability was ensured through SHAP analysis, revealing region type, maternal age, and administrative division as the most influential predictors and highlighting substantial geographic and demographic disparities. An interactive Tableau dashboard translates predictive insights into accessible visual analytics for policy and decision support. The findings underscore the importance of explicit imbalance handling and explainable artificial intelligence frameworks in maternal health risk modeling, offering a scalable and transparent approach for data-driven maternal mortality reduction strategies.
SN  - 3068-7403
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Mazumder2026Predictive,
  author = {Partha Pratim Mazumder and Raza Hasan and Salman Mahmood and Sellappan Palaniappan},
  title = {Predictive Analytics for Maternal Mortality in Bangladesh: An Interpretable ML Framework with Ensemble Methods},
  journal = {ICCK Transactions on Machine Intelligence},
  year = {2026},
  volume = {2},
  number = {3},
  pages = {127-143},
  doi = {10.62762/TMI.2026.182317},
  url = {https://www.icck.org/article/abs/TMI.2026.182317},
  abstract = {Maternal mortality in Bangladesh remains a critical public health challenge, with recent evidence indicating stagnation in mortality reduction despite expanded facility-based delivery and skilled birth attendance. Accurate identification of high-risk cases is essential to enable targeted intervention and resource allocation. This study develops an interpretable machine learning framework for maternal mortality prediction using the nationally representative Bangladesh Maternal Mortality Survey 2016 (BMMS-2016). A comprehensive data integration and feature engineering pipeline was implemented across demographic, socioeconomic, and maternal healthcare domains. Given the severe class imbalance inherent in maternal death outcomes, multiple resampling strategies---including Tomek Links undersampling, SMOTE, ADASYN, and CTGAN---were systematically evaluated in conjunction with diverse classifiers and ensemble methods. Among all configurations, Random Forest combined with SMOTE achieved the best overall performance (ROC-AUC: 0.9635; Accuracy: 0.9016; F1-score: 0.8928), demonstrating superior precision--recall balance suitable for rare-event prediction. Tree-based ensemble models consistently outperformed baseline classifiers. Model interpretability was ensured through SHAP analysis, revealing region type, maternal age, and administrative division as the most influential predictors and highlighting substantial geographic and demographic disparities. An interactive Tableau dashboard translates predictive insights into accessible visual analytics for policy and decision support. The findings underscore the importance of explicit imbalance handling and explainable artificial intelligence frameworks in maternal health risk modeling, offering a scalable and transparent approach for data-driven maternal mortality reduction strategies.},
  keywords = {maternal mortality prediction, interpretable machine learning, class imbalance techniques, SMOTE, ADASYN, CTGAN, Tomek Links, ensemble learning, Random Forest, SHAP explainability, healthcare risk modeling, public health analytics},
  issn = {3068-7403},
  publisher = {Institute of Central Computation and Knowledge}
}