Machine learning-based diabetes risk classification using clinical and lifestyle indicators

Diabetes mellitus is a major global health challenge associated with substantial morbidity, mortality, and healthcare costs. This study presents an internally validated machine-learning framework for multiclass classification of current diabetes-risk strata using concurrent clinical and lifestyle indicators. The dataset comprised 825 observations with nine predictors: age, gender, body mass index (BMI), systolic and diastolic blood pressure, glycated hemoglobin (HbA1c), fasting blood sugar (FBS), physical activity level, and family history of diabetes. Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP) classifiers were evaluated using an 80:20 stratified train-test split, stratified ten-fold cross-validation, accuracy, precision, recall, F1-score, and ROC-AUC. Model differences were assessed using the Friedman test followed by Holm-adjusted pairwise comparisons. XGBoost achieved the highest internal test performance, with an accuracy of 98.18%, precision of 98.22%, recall of 98.18%, F1-score of 98.18%, and ROC-AUC of 0.99970. Cross-validation showed stable internal performance, and the Friedman test indicated an overall difference among classifiers (χ² = 10.1786, p = 0.0171), although Holm-adjusted comparisons showed that XGBoost differed significantly from SVM but not from RF or MLP. Global feature-importance analysis identified FBS, systolic blood pressure, BMI, and HbA1c as the most influential predictors within the fitted models. Because the study used a modest, single-center cohort and did not include external validation, the findings should be interpreted as proof-of-concept evidence for classifying existing, source-defined diabetes-risk strata rather than predicting future diabetes onset or establishing clinical readiness. Independent multicenter validation is required before clinical implementation.

Authors

Institutions

Publication Details

Journal
Scientific Reports
Published
2026-11-14
DOI
https://doi.org/10.1038/s41598-026-78291-x
Primary Topic
Artificial Intelligence in Healthcare
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Machine learning-based diabetes risk classification using clinical and lifestyle indicators

Azhin Mohammed Khudhur, Dler Hussein Kadir, Hindreen Abdullah Taher, Sangar Anwar karim
Scientific Reports
Artificial Intelligence in Healthcare
article

Machine learning-based diabetes risk classification using clinical and lifestyle indicators

Azhin Mohammed Khudhur, Dler Hussein Kadir, Hindreen Abdullah Taher, Sangar Anwar karim
article en

Abstract

Diabetes mellitus is a major global health challenge associated with substantial morbidity, mortality, and healthcare costs. This study presents an internally validated machine-learning framework for multiclass classification of current diabetes-risk strata using concurrent clinical and lifestyle indicators. The dataset comprised 825 observations with nine predictors: age, gender, body mass index (BMI), systolic and diastolic blood pressure, glycated hemoglobin (HbA1c), fasting blood sugar (FBS), physical activity level, and family history of diabetes. Support Vector Machine (SVM), Random Forest (RF), Extreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP) classifiers were evaluated using an 80:20 stratified train-test split, stratified ten-fold cross-validation, accuracy, precision, recall, F1-score, and ROC-AUC. Model differences were assessed using the Friedman test followed by Holm-adjusted pairwise comparisons. XGBoost achieved the highest internal test performance, with an accuracy of 98.18%, precision of 98.22%, recall of 98.18%, F1-score of 98.18%, and ROC-AUC of 0.99970. Cross-validation showed stable internal performance, and the Friedman test indicated an overall difference among classifiers (χ² = 10.1786, p = 0.0171), although Holm-adjusted comparisons showed that XGBoost differed significantly from SVM but not from RF or MLP. Global feature-importance analysis identified FBS, systolic blood pressure, BMI, and HbA1c as the most influential predictors within the fitted models. Because the study used a modest, single-center cohort and did not include external validation, the findings should be interpreted as proof-of-concept evidence for classifying existing, source-defined diabetes-risk strata rather than predicting future diabetes onset or establishing clinical readiness. Independent multicenter validation is required before clinical implementation.

Scientific Reports
Cihan University-Erbil (IQ), Qaiwan International University (IQ), Salahaddin University-Erbil (IQ), University of Sulaimani (IQ)
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