Publication

Abstract : Depression being a prevalent mental health issue that significantly affects and impacts society, it calls for effective treatments that offer precise and timely prediction approaches. This paper introduces a unique hybrid model that combines Support Vector Machine (SVM) classification with Convolutional Neural Networks (CNNs). Due to potential risks associated with generating synthetic samples for medical datasets, the study evaluates the models both with and without the Synthetic Minority Over-sampling Technique (SMOTE). The study assessed K-Nearest Neighbors (KNN), Extreme Gradient Boosting (XGB), and SVM as baseline models, with SVM achieving the highest test accuracy among the three, recording 92.4% with SMOTE and 90% without SMOTE. The hybrid model, combining CNN and SVM with SMOTE augmentation, achieved a remarkable test accuracy of 93%, compared to 88.4% without SMOTE. These findings highlight the hybrid model's efficacy and stress how crucial it is to resolve class imbalance in medical datasets in order to improve forecast accuracy.