Text-Based Data Analysis for Mental Health Using Explainable AI and Deep Learning

This paper analyzes mental health through text-based data using explainable AI and deep learning techniques, focusing on the classification of addiction, alcoholism, anxiety, depression, and suicidal thoughts. The motivation for this research stems from the lack of suitable datasets from Twitter and the need for a more comprehensive understanding of mental health using text data. We propose a methodology that involves leveraging Reddit data, employing various text vectorization techniques (TF-IDF, Word2Vec, and GloVe), and implementing multiple classification algorithms (XGBoost, Decision Tree, SVM, Naive Bayes, Simple Gradient Descent, Stochastic Gradient Descent, K-Nearest Centroid, K-Nearest Neighbor, AdaBoost, Random Forest, and Logistic Regression) using the TF-IDF text vectorizer. The problem revolves around accurately classifying mental health-related topics based on text data. By employing a range of classification algorithms and text vectorization techniques, we aim to develop machine-learning models that effectively identify addiction, alcoholism, anxiety, depression, and suicidal thoughts within text-based discussions. This research contributes to mental health analysis using text-based data and advanced machine-learning techniques. The results highlight the potential of Reddit as a valuable resource for understanding mental health concerns and demonstrate the effectiveness of the proposed methodology. Our findings have far-reaching implications for mental health practitioners, researchers, and policymakers, facilitating the development of tailored interventions and support systems for those in need.