BXCNN: A novel depression detection model

Abstract

In response to the mounting societal pressures, an increasing number of individuals grappling with mental health challenges are turning to social media platforms to express their feelings. The utilization of deep learning models for analyzing social media data has become increasingly crucial in detecting early signs of depression. Early intervention through depression detection can significantly enhance patients quality of life and even save lives. However, many existing deep learning models suffer from low prediction accuracy, exacerbated by the imbalance between positive and negative samples in the collected data. To address these challenges, we propose a novel depression detection model integrated with BERT, XGBoost, and Convolutional Neural Networks (BXCNN). This model harnesses the advantages of ensemble learning and deep learning technologies by integrating XGBoost for feature extraction to alleviate data imbalance and CNN for classification. We transform depression-related textual data into sentence vectors using BERT to capture semantic information effectively. These features are then fed into a CNN classifier to accurately predict the likelihood of individuals exhibiting depressive symptoms. Through empirical evaluations on relevant datasets, our approach excels across various evaluation metrics, including accuracy, precision, recall, F1 score, and the area under the receiver operating characteristic curve (AUC-ROC).