Diabetes Prediction Model Based on SVM Optimized with RF Feature Selection and GWO

As the country with the heaviest diabetes burden, China faces significant impacts on its economic output and public health. Screening and predicting diabetes risk populations through machine learning holds practical significance. In this study, a random forest (RF) feature selection model was employed to filter a collected diabetes-related dataset, retaining eight high-importance features: glycated hemoglobin (HbA1c), fasting blood sugar (FBS), hypertension, systolic blood pressure (SBP), family history of diabetes, diet quality, diastolic blood pressure (DBP), and BMI. Among these, HbA1c and FBS exhibited significantly higher importance scores of 12.4429 and 10.7711, respectively. Subsequently, an SVM model optimized with the Grey Wolf Optimizer (GWO) was constructed based on the RF-filtered dataset and compared with PSO-SVM, BP neural network, and basic neural network models. The evaluation results were as follows: <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$A_{c c}=0.917, \text{AR}=0.907, \text{AP}=0.875, \mathrm{F} 1$</tex>-score <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$=0.891$</tex>, outperforming other models. This indicates that the proposed model demonstrates strong performance and potential in diabetes prediction.