Detection of Type 2 Diabetes Based on Electronic Health Records Using Hybrid Support Vector Machine & Random Forest

Effective Type 2 Diabetes treatment and patient follow-up depend on an early diagnosis. Machine Learning (ML) techniques have advanced with the increasing the amount of Electronic Health Record (EHR) data that is available. However, overfitting, interpretability issues, and computational costs could occur from managing and modeling large data sets. A study that focused on elements of the EHR collected from a homogeneous group of participants before the Type 2 diabetes (T2D) diagnosis using Support Vector Machine-Random Forest (SVM-RF) to identify type 2 diabetes. Compared to the previous machine learning methods, the proposed approach is more reliable, provides the best prediction performance-computation and improves the model's interpretability.