Personalized insulin dosing for Type 1 diabetes mellitus (T1DM) remains challenging because of complex glucose-insulin dynamics and substantial patient variability. Reinforcement learning (RL) has emerged as a promising approach for adaptive insulin management, yet the reliability of learned policies depends heavily on reward design and evaluation strategy. This study compares three actor–critic RL algorithms: Soft Actor-Critic (SAC), Advantage Actor-Critic (A2C), and Proximal Policy Optimization (PPO) for personalized insulin dosing using real-world continuous glucose monitoring, insulin delivery, basal insulin, and meal intake data from the OhioT1DM dataset. A custom Gymnasium-based environment was developed, and all algorithms were trained under identical conditions for 100,000 timesteps. Performance was evaluated using cumulative reward together with clinically relevant measures, including Time in Range (TIR) and insulin dosing behaviour. Although A2C and PPO achieved higher cumulative rewards than SAC, both converged to near-zero insulin dosing policies that exploited the reward formulation rather than learning clinically meaningful glucose regulation. In contrast, SAC maintained adaptive dosing behaviour, achieving a TIR of 72.71% with an average insulin dose of 1.769 U/step. These findings show that higher cumulative reward does not necessarily correspond to better clinical decision-making in open-loop reinforcement learning environments. The study highlights the importance of behaviour-focused evaluation alongside conventional reward metrics and provides practical insights for developing safer and more reliable reinforcement learning systems for personalized diabetes management.
C. M. Anyanwu, Nkiru C. Ogbonna, Mary Ofuru Kam et al.· International Journal of Lat...· 0 citations
Hypertension is one of the most important modifiable risk factors for Cardiovascular Disease (CVD), yet identifying which hypertensive patients are at higher risk remains challenging in clinical practice. This study developed and evaluated three machine-learning models: logistic regression, random forest, and Gradient Boosting for CVD risk prediction in a cohort of 23,543 hypertensive patients drawn from a 70,000 patient cardiovascular dataset. After preprocessing, feature engineering, SMOTE-based class balancing, and hyperparameter tuning via randomized search, model performance was assessed on a held-out test set and validated using 5-fold stratified cross-validation with SMOTE correctly nested inside each fold to avoid data leakage. On the test set, tuned Gradient Boosting model achieved the highest accuracy (78.59%) and AUC-ROC (0.6681), outperforming Logistic Regression (0.6633) and Random Forest (0.6508). cross-validation provided a slightly different perspective: Logistic Regression’s mean AUC-ROC (0.6628) edged out Gradient Boosting (0,6609) and Random Forest (0.6383), SHAP analysis on the Gradient Boosting model identified systolic blood pressure, age, and height as the strongest predictors, with height rivaling systolic blood pressure and surpassing BMI a notable difference from Random Forest’s feature importance ranking. Lifestyle factors (smoking, alcohol, physical activity) contributed minimally. These findings highlight blood pressure and body size measures as the dominant clinical signals in this dataset, while demonstrating the potential of an explainable machine-learning model based on routinely collected clinical data to support cardiovascular risk stratification and clinical decision-making in hypertensive patients, despite their moderate discriminative performance.
C. M. Anyanwu, J. C. Onyianta, Ogechi Gift Onyedi et al.· Nature Journal of Emerging S...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
