Selected work

Representative Papers

Explainable AI for Chronic Kidney Disease Prediction Using Simulated Federated Learning

Jul 28, 2026

This study addresses the challenges of early diagnosis of chronic kidney disease (CKD) and the limitations imposed by data silos and privacy constraints in multi-institutional healthcare settings. To overcome these issues, the authors propose an interpretable federated learning framework that integrates explainable artificial intelligence (XAI) with a voting-based ensemble strategy. The approach combines Random Forest, AdaBoost, and XGBoost into a VotingClassifier and employs GridSearchCV for hyperparameter optimization, enabling collaborative model training across institutions without compromising patient privacy. Experimental results demonstrate that the resulting global model achieves an average accuracy of 99%, substantially enhancing early diagnostic performance. This work represents the first successful implementation of a CKD prediction system that simultaneously ensures privacy preservation and model interpretability, thereby validating the feasibility and advantages of interpretable federated learning in clinical applications.

0 citationsRead paper

Spatial Prediction of Soil Microplastics and Organic Matter Using Graph Attention Networks

Jul 24, 2026

This study addresses the critical need for accurate assessment of ecosystem health and sustainable land use by pioneering the application of Graph Attention Networks (GAT) to spatial prediction of soil microplastics and organic matter. A two-layer GAT model was developed, integrating multisource environmental variables—including spatial coordinates, soil properties, and land use—to effectively capture local spatial dependencies among 91 georeferenced samples. The model achieved excellent predictive performance, with R² = 0.87 (RMSE = 625.06) for microplastics and R² = 0.91 (RMSE = 0.43) for organic matter, thereby demonstrating the substantial innovation potential and practical utility of GAT in soil spatial modeling.

0 citationsRead paper

Integrating Machine Learning Ensembles and Large Language Models for Heart Disease Prediction Using Voting Fusion

Feb 25, 2026

This study addresses the limited reliability of single-model approaches in early cardiovascular disease prediction by proposing a hybrid architecture that integrates ensemble learning with large language models (LLMs). The framework combines mainstream ensemble methods—XGBoost, LightGBM, CatBoost, and Random Forest—with open-source LLMs such as Gemini 2.5 Flash through a voting mechanism, leveraging the OpenRouter API to enable zero-shot and few-shot inference. Evaluated under uncertainty, the proposed method achieves a prediction accuracy of 96.62% and an AUC of 0.97, substantially outperforming standalone ensemble models (95.78%) and LLMs alone (maximum 78.9%). This work presents the first empirical validation of an ML–LLM voting fusion strategy for clinical decision support, demonstrating its enhanced robustness and efficacy in real-world diagnostic scenarios.

0 citationsRead paper

Recent publications

Latest Papers

Explainable AI for Chronic Kidney Disease Prediction Using Simulated Federated Learning

Jul 28, 2026

This study addresses the challenges of early diagnosis of chronic kidney disease (CKD) and the limitations imposed by data silos and privacy constraints in multi-institutional healthcare settings. To overcome these issues, the authors propose an interpretable federated learning framework that integrates explainable artificial intelligence (XAI) with a voting-based ensemble strategy. The approach combines Random Forest, AdaBoost, and XGBoost into a VotingClassifier and employs GridSearchCV for hyperparameter optimization, enabling collaborative model training across institutions without compromising patient privacy. Experimental results demonstrate that the resulting global model achieves an average accuracy of 99%, substantially enhancing early diagnostic performance. This work represents the first successful implementation of a CKD prediction system that simultaneously ensures privacy preservation and model interpretability, thereby validating the feasibility and advantages of interpretable federated learning in clinical applications.

0 citationsRead paper

Spatial Prediction of Soil Microplastics and Organic Matter Using Graph Attention Networks

Jul 24, 2026

This study addresses the critical need for accurate assessment of ecosystem health and sustainable land use by pioneering the application of Graph Attention Networks (GAT) to spatial prediction of soil microplastics and organic matter. A two-layer GAT model was developed, integrating multisource environmental variables—including spatial coordinates, soil properties, and land use—to effectively capture local spatial dependencies among 91 georeferenced samples. The model achieved excellent predictive performance, with R² = 0.87 (RMSE = 625.06) for microplastics and R² = 0.91 (RMSE = 0.43) for organic matter, thereby demonstrating the substantial innovation potential and practical utility of GAT in soil spatial modeling.

0 citationsRead paper

Integrating Machine Learning Ensembles and Large Language Models for Heart Disease Prediction Using Voting Fusion

Feb 25, 2026

This study addresses the limited reliability of single-model approaches in early cardiovascular disease prediction by proposing a hybrid architecture that integrates ensemble learning with large language models (LLMs). The framework combines mainstream ensemble methods—XGBoost, LightGBM, CatBoost, and Random Forest—with open-source LLMs such as Gemini 2.5 Flash through a voting mechanism, leveraging the OpenRouter API to enable zero-shot and few-shot inference. Evaluated under uncertainty, the proposed method achieves a prediction accuracy of 96.62% and an AUC of 0.97, substantially outperforming standalone ensemble models (95.78%) and LLMs alone (maximum 78.9%). This work presents the first empirical validation of an ML–LLM voting fusion strategy for clinical decision support, demonstrating its enhanced robustness and efficacy in real-world diagnostic scenarios.

0 citationsRead paper