Early Detection of Mental Health Disorders based on Sentiment using Stacking Method

Abstract

Mental health disorders are a serious and growing global concern, including in Indonesia. This study aims to predict mental health disorders through sentiment analysis using the Stacking Classifier approach, which combines Random Forest, Gradient Boosting Classifier, and Logistic Regression algorithms. The dataset was sourced from various social media platforms, consisting of textual data classified into seven mental health categories, such as depression, anxiety, and personality disorders. The data underwent preprocessing steps, including cleaning, balancing, and dimensionality reduction using the TF-IDF algorithm. The study results indicate that the Stacking Classifier method achieved an accuracy of 95.66%, with a precision of 95.63%, recall of 95.66%, and F1-Score of 95.64%. These results outperform the individual algorithms tested in the research. The findings demonstrate the significant potential of sentiment analysis powered by machine learning for early detection of mental health disorders, making it a valuable tool to enhance diagnosis and intervention in mental health care more effectively.

Keywords

• mental health, sentiment analysis, stacking classifier, machine learning, mental health prediction.