Classification of Mental Health Conditions in Reddit Post using Multinomial Naïve Bayes Algorithm

Mental health plays a pivotal role in well-being, yet mental health challenges such as depression, anxiety, and stress (DAS) often remain underdiagnosed or misclassified. This paper introduces a novel chatbot-based classification system leveraging the Multinomial Naïve Bayes algorithm to analyze Red...

詳細記述

書誌詳細
出版年:2024 IEEE 22nd Student Conference on Research and Development, SCOReD 2024
第一著者: 2-s2.0-85219584809
フォーマット: Conference paper
言語:English
出版事項: Institute of Electrical and Electronics Engineers Inc. 2024
オンライン・アクセス:https://www.scopus.com/inward/record.uri?eid=2-s2.0-85219584809&doi=10.1109%2fSCOReD64708.2024.10872671&partnerID=40&md5=6d8661132c873dc001c305cfd362c9af
その他の書誌記述
要約:Mental health plays a pivotal role in well-being, yet mental health challenges such as depression, anxiety, and stress (DAS) often remain underdiagnosed or misclassified. This paper introduces a novel chatbot-based classification system leveraging the Multinomial Naïve Bayes algorithm to analyze Reddit posts and assess users' mental health conditions. Using the DASS-21 questionnaire as a framework, the chatbot converts traditional multiple-choice queries into open-ended questions, allowing for richer expression and deeper insight into user emotions. The project achieved an improved classification accuracy of 83% following hyperparameter tuning and K-fold cross-validation. Exploratory data analysis, including word cloud visualizations, revealed recurring themes associated with DAS, highlighting the system's interpretative capabilities. The chatbot's seamless integration into a web and mobile application further enhances its accessibility and user engagement. While demonstrating significant progress in mental health classification, the research acknowledges limitations in dataset diversity and conversational capabilities. Future work aims to expand coverage to additional mental health conditions and improve mental health detection. © 2024 IEEE.
ISSN:
DOI:10.1109/SCOReD64708.2024.10872671