Comparative Analysis of Machine Learning and Deep Learning Algorithms for Twitter-Based Depression Detection

Título: Comparative Analysis of Machine Learning and Deep Learning Algorithms for Twitter-Based Depression Detection

Autores: Emanoel Faria dos Santos, Lucas Grassano Lattari, Maurício Archanjo Nunes Coelho, Bianca Portes de Castro

Resumo: Depression is a significant mental health disorder affecting millions of individuals globally. Detecting depressive symptoms from written texts, especially on social media platforms like Twitter, has received considerable attention. In this paper, we present a comparative analysis of machine learning and deep learning algorithms for depression detection on Twitter. We propose an innovative approach that integrates a multilayer Long Short-Term Memory (LSTM) architecture with a Multi-Head Attention component. Our approach achieves up to 99% across all key metrics, including accuracy, recall, F1- score, and precision. However, it should be noted that these high scores are obtained in certain instances, thus being highly competitive compared to other relevant works. Despite facing challenges such as imbalanced datasets and user-annotated data, these remarkable results mark a promising advancement in the field of text-based depression detection

Palavras-chave: depression detection, twitter, social media, machine learning, deep learning, sentiment analysis, lstm, multihead attention

Páginas: 7

Código DOI: 10.21528/CBIC2023-061

Artigo em pdf: CBIC_2023_paper061.pdf

Arquivo BibTeX: CBIC_2023_061.bib