Construction of a diabetic foot prediction model based on particle swarm optimization and interpretable algorithms
Abstract
ObjectiveTo analyze the influencing factors of diabetic foot and to establish a predictive model.MethodsClinical data from 551 inpatients with type 2 diabetes mellitus were retrospectively enrolled.Missing data were handled using multivariate imputation by chained equations(MICE).The dataset was divided into training and test sets at a 7∶3 ratio.Feature selection on the training set was performed using particle swarm optimization(PSO).The optimal feature subset was used to construct a random forest(RF) prediction model (PSO⁃RF).Four additional machine learning models,SVM,XGBoost,logistic regression,and decision tree,were concurrently developed as comparators.Model performance was evaluated using accuracy,precision,recall,F1 score,and area under the receiver operating characteristic curve(AUC).Key risk indicators were identified by combining XGBoost feature importance and the SHAP interpretability algorithm.ResultsThe PSO⁃RF model achieved an accuracy of 0.724,precision of 0.747,recall of 0.704,F1 score of 0.724,and AUC of 0.750,demonstrating superior predictive performance over the other four conventional classification models.Consistent results were obtained from PSO⁃based feature selection,XGBoost feature importance,and SHAP analysis, identifying plateletcrit,serum creatinine,and fasting blood glucose as core risk markers for diabetic foot.ConclusionsThe early prediction model for diabetic foot constructed using PSO⁃RF exhibits favorable predictive performance.Plateletcrit, creatinine,and fasting blood glucose can serve as key laboratory indicators for screening and stratified management of high⁃risk populations for diabetic foot.
Related
- Construction of an interpretable machine learning clinical prediction model for renal fibrosis based on renal-cortex ultrasound radiomics
- Construction and validation of an identification model for frailty in maintenance hemodialysis patients using multiple machine learning algorithms: a cross-sectional study of retrospectively collected data
- Interpretable Diabetic Retinopathy Classification Using Hybrid ConvNeXt–MLP-Mixer and ConvNeXt–PoolFormer Architectures
- Development and interpretable validation of a machine-learning model for edentulism risk in US adults : A cross-sectional study integrating five algorithms and dietary data from NHANES 2009–2014
- KT-CBHNET: a spatio-temporal deep learning framework for interpretable ECG signal classification for cardiovascular disease prediction
BibTeX
@article{jingmei2026construction,
title = {{Construction of a diabetic foot prediction model based on particle swarm optimization and interpretable algorithms}},
author = {SONG Jingmei and LI Xiaohong and PAN Yifeng and LI Gehong},
journal = {Huli yanjiu},
year = {2026},
month = oct,
url = {http://hlyj.suo1.cn/thesisDetails#10.12102/j.issn.1009-6493.2026.21.012},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Construction of a diabetic foot prediction model based on particle swarm optimization and interpretable algorithms
AU - SONG Jingmei
AU - LI Xiaohong
AU - PAN Yifeng
AU - LI Gehong
PY - 2026
DA - 2026/10/30
JO - Huli yanjiu
UR - http://hlyj.suo1.cn/thesisDetails#10.12102/j.issn.1009-6493.2026.21.012
AB - ObjectiveTo analyze the influencing factors of diabetic foot and to establish a predictive model.MethodsClinical data from 551 inpatients with type 2 diabetes mellitus were retrospectively enrolled.Missing data were handled using multivariate imputation by chained equations(MICE).The dataset was divided into training and test sets at a 7∶3 ratio.Feature selection on the training set was performed using particle swarm optimization(PSO).The optimal feature subset was used to construct a random forest(RF) prediction model (PSO⁃RF).Four additional machine learning models,SVM,XGBoost,logistic regression,and decision tree,were concurrently developed as comparators.Model performance was evaluated using accuracy,precision,recall,F1 score,and area under the receiver operating characteristic curve(AUC).Key risk indicators were identified by combining XGBoost feature importance and the SHAP interpretability algorithm.ResultsThe PSO⁃RF model achieved an accuracy of 0.724,precision of 0.747,recall of 0.704,F1 score of 0.724,and AUC of 0.750,demonstrating superior predictive performance over the other four conventional classification models.Consistent results were obtained from PSO⁃based feature selection,XGBoost feature importance,and SHAP analysis, identifying plateletcrit,serum creatinine,and fasting blood glucose as core risk markers for diabetic foot.ConclusionsThe early prediction model for diabetic foot constructed using PSO⁃RF exhibits favorable predictive performance.Plateletcrit, creatinine,and fasting blood glucose can serve as key laboratory indicators for screening and stratified management of high⁃risk populations for diabetic foot.
SN - 1009-6493
ER -