Uso de aprendizado de máquina para identificação de perfis depressivos em redes sociais

Abstract

Regardless of age, social class or nationality, depression is a psychic disorder that affects millions of people in the world. Sentiment Analysis has been used to extract relevant information from social media users’ posts in order to identify their feelings regarding the most varied subjects. The present study aims to use Sentiment Analysis techniques to detect depressive profiles on Twitter. To this end, a strategy entitled DP-Symptom-Identifierwas created which identifies possible depressive symptoms present in tweets and then enables profile analysis as a whole, temporally, facilitating the process of identifying depressive profiles in timelines. Experiments were conducted to identify the best machine learning algorithm to compose the strategy and also to evaluate the efficiency of the DP-Symptom-Identifierstrategy. The algorithm selected was the Multilayer Perceptron, which obtained a 90% assertiveness rate with a processing time of fewer than 5 seconds. The proposed strategy proved effective in identifying possible depressive profiles analyzing timelines extracted from Twitter. Therefore, the results indicate that the strategy can identify symptoms of depression and allows the identification of depressive profiles correctly.