PROJECT TITLE :

Discovering the Type 2 Diabetes in Electronic Health Records Using the Sparse Balanced Support Vector Machine

ABSTRACT:

The early detection of type 2 diabetes (T2D) is critical for an effective T2D integrated management strategy and patient follow-up. In recent years, there has been a significant increase in the amount of available electronic health record (EHR) data, and Machine Learning (ML) techniques have evolved significantly. However, organizing and modeling such a large amount of data can present a number of problems, including overfitting, model interpretability, and computational expense. We developed a machine learning method called sparse balanced support vector machine (SB-SVM) to detect T2D in a newly gathered EHR dataset based on these objectives (named Federazione Italiana Medici di Medicina Generale dataset). We chose only those obtained before T2D diagnosis from a uniform age range of individuals, out of all the EHR features pertaining to exemptions, examinations, and drug prescriptions. We demonstrated the reliability of the proposed strategy in comparison to other machine learning and Deep Learning approaches commonly used in the state-of-the-art to solve this problem. The SB-SVM outperforms other state-of-the-art competitors in terms of predicting performance and computation time, according to the results. Furthermore, the induced sparsity improves model interpretability by automatically managing high-dimensional data and the common imbalanced class distribution.