AI-Driven Mortality Risk Classification Using HMIS Data: A Case Study from Nagaon District, Assam, India

Abstract

Maternal, infant, and adult mortality is one of the vital public health concerns in India. In the present study machine learning models were implemented on Health Management Information System (HMIS) dataset, considering Nagaon district of Assam, India for the classification of district blocks to low and high-mortality risk groups. Various service indicators and Mortality were accumulated at the block level. Different classifiers like, Random Forest, Logistic Regression, Linear and RBF SVM, KNN, and Gaussian Naive Bayes were evaluated. Repeated 5×5 Cross-validation technique is used for the analysis and evaluated purpose. Random Forest (RF) produce an accuracy of 86.7%. Out of the total classifiers applied Random Forest turned out to be the best-balanced technique. While in terms of ROC-AUC, RBF SVM generated highest 96%, which signifying robust discriminative capacity. Logistic Regression and SVM remain stable and explainable baselines. KNN and Naive Bayes yield modest performance. Analysis confirmed that antenatal care and child immunization as the vital predictors of mortality consequences. Moreover, these outcomes highlight the necessity of robust justification in AI-driven system for health analytics along with validate the potential of ensemble method and SVM technique to support HMIS-based mortality investigation and guide aimed interventions. From a public health viewpoint, these outcomes advocate that AI-driven analytics can improve routine HMIS data use for evidence-based planning. Recognizing the high-risk mortality blocks, respective department of health can line up interventions viz., antenatal care, immunization, and maternal health services.

Ähnliche Arbeiten