MO-AIWOA-XAI: A Multi-Objective Explainable Diabetes Risk Prediction integrating Feature Optimization and Cross-Dataset Validation

Abstract Clinical decision-making for diabetes management requires prediction models that are both accurate and interpretable. Although machine learning techniques have demonstrated promising performance in diabetes risk prediction, their adoption in healthcare is often constrained by limited transparency and feature redundancy. To address these challenges, the study proposes a Multi-Objective Adaptive Improved Whale Optimization Explainable Artificial Intelligence (MO-AIWOA-XAI) framework for diabetes risk prediction. The proposed framework integrates a multi-objective Adaptive Improved Whale Optimization Algorithm (AIWOA) for feature selection, an Optuna-optimized Extra Trees classifier, and explainable artificial intelligence techniques including SHAP, LIME, and DiCE. The feature selection strategy simultaneously maximizes predictive performance while minimizing feature dimensionality, enabling the identification of clinically relevant risk factors. The framework was evaluated using the PIMA Indians Diabetes Dataset and the Early Stage Diabetes Risk Prediction Dataset. Data preprocessing involved KNN imputation, feature engineering, Min-Max normalization, outlier treatment, and SMOTE-Tomek class balancing. Extensive experiments, including comparisons with PSO, GA, and WOA, cross-dataset validation, 10-fold cross-validation, calibration analysis, and decision curve analysis, were performed to assess model robustness. The proposed framework achieved an accuracy of 83.76%, AUC of 0.9004, F1-score of 0.84, MCC of 0.6757, and PR-AUC of 0.9068 on the PIMA dataset, while attaining 95.19% accuracy and 0.992 AUC on the Early Stage Diabetes dataset. In addition, the framework produced stable and clinically interpretable explanations through SHAP, LIME, and counterfactual analysis. The results demonstrate that the proposed MO-AIWOA-XAI framework effectively combines feature optimization, predictive modeling, and explainable artificial intelligence to provide a reliable and clinically interpretable solution for diabetes risk assessment and decision support.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-12
DOI
https://doi.org/10.21203/rs.3.rs-10356229/v1
Primary Topic
Artificial Intelligence in Healthcare
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

MO-AIWOA-XAI: A Multi-Objective Explainable Diabetes Risk Prediction integrating Feature Optimization and Cross-Dataset Validation

krishnaprakasha k, Sumalatha U
Research Square
Artificial Intelligence in Healthcare
preprint

MO-AIWOA-XAI: A Multi-Objective Explainable Diabetes Risk Prediction integrating Feature Optimization and Cross-Dataset Validation

krishnaprakasha k, Sumalatha U
preprint en

Abstract

Abstract Clinical decision-making for diabetes management requires prediction models that are both accurate and interpretable. Although machine learning techniques have demonstrated promising performance in diabetes risk prediction, their adoption in healthcare is often constrained by limited transparency and feature redundancy. To address these challenges, the study proposes a Multi-Objective Adaptive Improved Whale Optimization Explainable Artificial Intelligence (MO-AIWOA-XAI) framework for diabetes risk prediction. The proposed framework integrates a multi-objective Adaptive Improved Whale Optimization Algorithm (AIWOA) for feature selection, an Optuna-optimized Extra Trees classifier, and explainable artificial intelligence techniques including SHAP, LIME, and DiCE. The feature selection strategy simultaneously maximizes predictive performance while minimizing feature dimensionality, enabling the identification of clinically relevant risk factors. The framework was evaluated using the PIMA Indians Diabetes Dataset and the Early Stage Diabetes Risk Prediction Dataset. Data preprocessing involved KNN imputation, feature engineering, Min-Max normalization, outlier treatment, and SMOTE-Tomek class balancing. Extensive experiments, including comparisons with PSO, GA, and WOA, cross-dataset validation, 10-fold cross-validation, calibration analysis, and decision curve analysis, were performed to assess model robustness. The proposed framework achieved an accuracy of 83.76%, AUC of 0.9004, F1-score of 0.84, MCC of 0.6757, and PR-AUC of 0.9068 on the PIMA dataset, while attaining 95.19% accuracy and 0.992 AUC on the Early Stage Diabetes dataset. In addition, the framework produced stable and clinically interpretable explanations through SHAP, LIME, and counterfactual analysis. The results demonstrate that the proposed MO-AIWOA-XAI framework effectively combines feature optimization, predictive modeling, and explainable artificial intelligence to provide a reliable and clinically interpretable solution for diabetes risk assessment and decision support.

Research Square
Manipal Academy of Higher Education (IN)
Good health and well-being
Artificial Intelligence in Healthcare
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon