Adolescents are at high risk for mental health issues, including depression, due to factors such as social media exposure, peer pressure, and behavioral influences. Early detection of at-risk individuals enables timely intervention and support. This study presents a hybrid depression detection model that integrates Naïve Bayes, TF-IDF, and emoji sentiment analysis to enhance classification accuracy. The approach utilizes textual and emoji-based sentiment cues extracted from social media data to identify depressive tendencies. The model is evaluated against traditional classifiers, demonstrating improved precision, recall, and F1-score. A comparative analysis of different machine learning techniques highlights the model’s effectiveness in detecting depression-related expressions. The findings emphasize the potential of AI-driven approaches in mental health monitoring, particularly for adolescents, by leveraging real-time social media interactions. This research contributes to the growing field of automated mental health assessment, offering a scalable and data-driven solution for early intervention. Future work will explore deep learning and multimodal analysis to further refine depression detection and provide more comprehensive insights into mental health patterns.
Introduction
Mental health disorders, especially depression, are increasingly prevalent, with social media serving as a key platform where people express emotions. Traditional clinical diagnosis is resource-heavy, prompting research into automated AI models for early depression detection from social media text.
Existing models mostly use text-based sentiment analysis with machine learning (e.g., SVM, deep learning), but often miss the nuanced emotional cues, especially those conveyed by emojis. This study proposes a hybrid approach combining Naïve Bayes classification, TF-IDF feature extraction, and emoji sentiment analysis to improve detection accuracy by capturing both linguistic and non-linguistic emotional cues.
The research contributions include:
A novel hybrid model integrating text and emoji sentiment for better depression detection.
Comparative evaluation showing improved accuracy over traditional methods.
Application to real-world social media datasets, ensuring robustness across diverse language use.
Advancing multimodal sentiment analysis for AI-driven mental health diagnostics.
Methodology:
Data collected from labeled depressive/non-depressive tweets.
Text preprocessing (normalization, tokenization, stopword removal, stemming) and emoji extraction with sentiment scoring.
Features engineered using TF-IDF and emoji sentiment scores.
Model trained with Multinomial Naïve Bayes classifier, combining text and emoji sentiment.
Evaluated using accuracy, precision, recall, and F1-score.
System & Results:
Web-based system built with Flask for user-friendly depression prediction from tweets.
Model achieved 90.38% accuracy, 0.90 precision, 0.89 recall, and 0.89 F1-score, demonstrating effective detection capability.
Conclusion
The proposed system provides an effective and scalable solution for analyzing tweets to predict depression by combining text-based sentiment analysis and symbol sentiment analysis. The system achieved an accuracy of 90.38% using a Multinomial Naive Bayes classifier, demonstrating its effectiveness in classifying tweets as \"Depressed\" or \"Not Depressed.\" The custom symbol sentiment mapping allowed the system to capture emotional cues from symbols, further improving the accuracy of predictions.
References
[1] Urvashi Panchal, Gonzalo Salazar de Pablo, Macarena Franco. \"The impact of COVID 19 lockdown on child and adolescent mental health: systematic review\". European Child & Adolescent Psychiatry (2023).
[2] Margarita Rodríguez and Antonio Casanez-Ventura. \"A review on sentiment analysis from social media platforms\". Expert Systems With Applications 223 (2023).
[3] Patti M. Valkenburg, Adrian Meier, and Ine Beyens. \"Social media use and its impact on adolescent mental health: An umbrella review of the evidence\". ScienceDirect, Current Opinion in Psychology 2022.
[4] Nicholas Pudjihartono, TayazaFadason, Andreas W. Kempa-Liehr, and Justin M. Sullivan. \"A Review of Feature Selection Methods for Machine Learning-Based Disease Risk Prediction\". Frontiers in Bioinformatics. Published: 27 June 2022.
[5] Qianwen Xu, Ariel Victor, Victor Chang, Christina Jayne. \"A systematic review of social media-based sentiment analysis: Emerging trends and challenges.\" Decision Analytics Journal, Volume 3, 2022.
[6] Untung Rahardja. \"Social Media Analysis as a Marketing Strategy in Online Marketing Business\". Startupreneur Business Digital (SABDA) Vol. 1 No. 2, October 2022.
[7] Amna Amanat, Muhammad Rizwan, Abdul Rehman Javed. \"Deep Learning for Depression Detection from Textual Data\". Electronics 2022, 11, 676. 23 February 2022.
[8] Nirmal Varghese Babu, E Grace Mary Kanaga. \"Sentiment Analysis in Social Media Data for Depression Detection Using Artificial Intelligence: A Review\". SN Computer Science (2022).
[9] Jeff Bostic, Sharon Hoover. \"Schools As a Vital Component of the Child and Adolescent Mental Health System\". Psychiatric Services 72:1, January 2021.
[10] Natasha R. Magson, Jasmine Fardouly, and Justin Y. A. Freeman. \"Risk and Protective Factors for Prospective Changes in Adolescent Mental Health during the COVID-19 Pandemic\". Journal of Youth and Adolescence (2021).
[11] Olympia L. K. Campbell, David Bann, and PraveethaPatalay. \"The gender gap in adolescent mental health: A cross-national investigation of 566,829 adolescents across 73 countries\". SSM-Population Health 13 (2021).
[12] Nikhil Kumar Singh, Deepak Singh Tomar, and Arun Kumar Sangaiah. \"Sentiment analysis: A review and comparative analysis over social media.\" Journal of Ambient Intelligence and Humanized Computing, 2018.