A SVM-based classification approach for obsessive compulsive disorder by oxidative stress biomarkers

Abstract

Current literature has the evidence of the role of oxidative stress in obsessive compulsive disorder (OCD). The present study aims to examine whether distinct patterns exist in the oxidative stress biomarkers (osbMarkers) in “Individuals with Obsessive Compulsive Disorder (OCDP)”, “their non-affected First-Degree Relatives (FDR)” and “Healthy-controls (HC)”. We suggest a theoretical and experimental study intended at determining the potential of classical classification approaches in finding patterns in osbMarkers such as Superoxide dismutase (SOD), Glutathione Peroxidase (GPX), Catalase (CAT), Malondialdehyde (MDA) and serum cortisol (COR)) for distinguishing OCDP, FDR and HC. A Support Vector Machine (SVM)-based approach has been used for classifying the three classes from osbMarkers. Radial kernel was used as kernel function in the proposed SVM approach. Optimum kernel parameters were obtained by using a grid search based method. For the present investigation, SVM with linear kernel (SVML), Random forest (RF), Linear discriminant analysis (LDA) and K-Nearest Neighbor (KNN) are used as the base classifier. Experiments are carried out on osbMarkers and the results are compared with the classification techniques using SVML, RF, LDA and KNN. Results show that the proposed work has a significant edge over the other techniques for osbMarkers classification.