Application of Support Vector Machine for Detection of Functional Limitations in the Diabetic Patients of the Northwest of IRAN in 2017: A Descriptive Study

Support vector machine (SVM) is a robust and effective statistical method for the diagnosis and prediction of clinical outcomes based on combinations of predictor variables. The aim of this study was to use SVM to detect the functional limitations in the diabetic patients and evaluate the accuracy of this diagnosis.

This descriptive study was conducted on 378 diabetic patients referred to the diabetic centers of Ardabil and Tabriz in 2014-2015. To classify the diabetic patients in terms of functional limitation, based on the demographic and clinical variables, SVM was used with RBF (radial basis function) kernel and the training and test validation method. Evaluation was performed based on diagnostic indices including sensitivity, specificity, accuracy and area under the ROC (receiver operating characteristic) curve.

The results of SVM method showed that the classification accuracy, sensitivity, specificity of the SVM method in differentiating and correct diagnosis of functional limitations in the diabetic patients were 99%, 100% and 97%, respectively. The area under the ROC curve as the detection performance analysis of this model was 0.98.

In this study, SVM was used to classify the functional limitation status of the diabetic patients, and the results showed that the model had an acceptable performance. Considering the importance of classifying the medical outcomes correctly based on the combinations of predictor variables, the use of the methods such as SVM that are able to find optimal combinations could be helpful.