CommentSense: Multilabel Social Support Classification of Filipino-English Endocrinology Facebook Comments Using Machine Learning Classification Models

Abstract

The mental health implications of Diabetes can be mitigated through the presence of a social support system, and social media platforms offer a convenient avenue for fostering such support, even among healthcare professionals and patients. This study introduces novel machine learning models tailored for multilabel social support classification of Filipino-English Endocrinology Facebook comments. The objective is to effectively categorize comments into distinct types of support, including informational, emotional, appraisal, instrumental, and spam, thereby enabling healthcare professionals to efficiently manage their social media groups. The dataset underwent manual data cleaning and was subsequently divided into training and testing sets. Preprocessing techniques encompassing lowercasing, tokenization, and TF-IDF vectorization were employed on both sets. To address dataset imbalances, data augmentation techniques were implemented. Notably, the LP-SVM model emerged as the top performer and was seamlessly integrated into the CommentSense application. These findings enhance our comprehension of social support dynamics and furnish practitioners with a user-friendly tool for social support text classification.