AI for Public Health: Interpretable Supervised Encoder and Stroke Recurrence Prediction
Date:
- Proposed a segmented neural network-driven and supervised encoder to predict stroke recurrence probability and interpret its risk factors
- Preserved the interpretability at the risk factor level, while improved the decision-making process together with widely used downstream classifiers such as logistic regression and tree-based models
- Achieved higher AUC (90+%) in both ROC and Precision-Recall curves compared to the benchmark, and consistently enhanced sensitivity and specificity in predicting stroke recurrence