International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 11 Issue: 10 | Oct 2024
p-ISSN: 2395-0072
www.irjet.net
LEVERAGING MACHINE LEARNING AND NLP FOR PERSONALIZED MENTAL HEALTH ANALYSIS FROM SOCIAL MEDIA INSIGHTS Dr. R. Lakshmi1, S. Ramya Sree2, B.J.A Rishi Priya3, R.P. Priyanka4 1Professor, Department of Computer Science and Engineering, K.L.N. College of Engineering, Sivagangai,
Tamil Nadu, India. 2,3,4,5 Final Year UG Students, Department of Computer Science and Engineering, K.L.N. College of Engineering,
Sivagangai, Tamil Nadu, India.
---------------------------------------------------------------------***--------------------------------------------------------------------Abstract - This study presents a machine learning (ML) Traditional mental health assessments, while effective, may lack the immediacy and nuance of real-time data. Social and natural language processing (NLP) framework for media offers an alternative, capturing natural language assessing mental health through social media text analysis. With the increasing amount of user-generated content, patterns that reflect users' daily experiences. Our dualsocial media provides a valuable source of data for model framework addresses two objectives: (1) binary classification to identify general mental health concerns and identifying mental health markers through expressed (2) multiclass classification for specific conditions like emotions, challenges, and general sentiment. This research depression, anxiety, and stress. aims to fulfill two primary objectives: (1) early detection of potential mental health concerns and (2) classification of Core techniques include sentiment analysis using VADER, specific conditions, including depression, anxiety, and stress, based on sentiment and linguistic patterns. Our dual-model feature extraction through TF-IDF, and dimensionality reduction via Singular Value Decomposition (SVD). We approach includes a binary classifier to detect general validate the framework using two datasets—one for mental health concerns and a multiclass classifier for general mental health and one for Twitter mental health— condition-specific categorization. The integration of VADER sentiment scoring, TF-IDF vectorization, and Truncated demonstrating high predictive accuracy across various mental health indicators. Singular Value Decomposition (SVD) for dimensionality reduction enables accurate feature extraction and enhanced The organization of this paper is as follows: Section II model performance. Experimental results on publicly details our methodology and preprocessing; Section III available datasets show high predictive accuracy, with presents experimental findings; Section IV discusses confusion matrix analysis validating minimal overlap across applications and limitations; and Section V concludes with classified mental health conditions. Potential applications future directions. include real-time mental health monitoring and tailored interventions, with future directions focused on multilingual 2.METHODOLOGY adaptation and expanded feature engineering. This framework provides a promising basis for automated, 2.1 Data Collection scalable, and personalized mental health support. Two public datasets were utilized: the general mental Key Words: Mental health assessment, social media health dataset and the Twitter mental health dataset. The analysis, machine learning, natural language former includes text data reflecting various mental health processing, sentiment analysis, VADER, Truncated SVD, conditions, while the latter captures real-time emotional binary classification, multiclass classification, expressions from users on social media platforms. personalized mental health support.
2.2 Data Preprocessing
1. INTRODUCTION
Text Cleaning: We removed special characters, URLs, and stop words from the text data to enhance the quality of the input for analysis.
Mental health significantly impacts quality of life and productivity, yet barriers like social stigma and limited resources hinder timely care. The extensive use of social media platforms provides a rich source of user-generated content, often revealing users' emotional states, mental challenges, and general well-being. This study leverages machine learning (ML) and natural language processing (NLP) to assess mental health from social media data, with a focus on early detection and classification of specific mental health conditions.
© 2024, IRJET
|
Impact Factor value: 8.315
Label Encoding: Target labels were encoded for both binary and multiclass classification tasks, facilitating model training.
|
ISO 9001:2008 Certified Journal
|
Page 761