Enhancing Mental Health Diagnosis with DA-TabSVM: A Multi-Class Hybrid Approach for Detecting Depression and Anxiety

Following the COVID-19 pandemic, as individuals adapted to a new normal, the likelihood of despair and anxiety rose. Depression’s effects can be lessened by early identification, suitable treatment, and support. The Sustainable Development Goals (SDGs) goals aim to reduce the burden of mental diseases by promoting mental health and well-being through comprehensive health strategies, reducing inequalities, and fostering supportive environments. Lately, automatic social media depression detection has gained attention as a promising field of study. However, we suggest a technique for using a model that learns to answer a mental health questionnaire and utilizing it to produce population-level predictions because there is a dearth of comprehensive annotated data. The purpose of this paper is to efficiently predict mental health conditions like depression and anxiety using data from psychological instruments. In this study, we propose an improved hybrid model employing DT+kNN and RF+NN which were applied on three datasets. Further, we propose a novel prediction model, DA-TabSVM, that can efficiently predict depression and anxiety using multiclass classification. The results show that the DT+kNN hybrid model achieves a maximum accuracy of 95% for predicting depression and 97% for predicting anxiety. Similarly, the RF+NN hybrid model achieves a maximum accuracy of 98% for predicting depression and anxiety. The performance of DA-TabSVM reached up to 97% in terms of accuracy on Dataset 1, 99% on Dataset 2, and 99% on Dataset 3 for depression detection and similarly high when detecting anxiety with a range reaching up to an accuracy of 100% on Dataset 3. Unlike standard strategies that may struggle with intricate feature relationships or be less flexible when dealing with different datasets, our hybrids make use of the complementary skills of many algorithms. Our findings demonstrate the usefulness of hybrid classifiers in mental health evaluation, advancing our knowledge in this crucial field.