Depression Detection on Social Media Twitter Using Long Short-Term Memory

Abstract

Mental health problems in the world, especially in Indonesia are still significant. According to the Ministry of Health of the Republic of Indonesia stated that depression is experienced by adolescents from the age of 15 to 24 years. The depression experienced by a person is sometimes not realized by the sufferer, so social media becomes an intermediary to express feelings in text form. From the available data, this case pushes the research to detect depression disorder. Detecting depression performs to know the Twitter user who experiences depression. Data used from 159 Twitter users for every username is taken from 100 tweets. In this research, we use Word2Vec and LSTM (Long Short-Term Memory) features as the classification method. The Word2Vec works in converting data as vector and seeing the relation for every word. LSTM is chosen since the dataset is used to collect tweet from the past tense and this method be able to save the data from the past doing prediction. The classification is performed by processing the data trained such as tweeting which becomes a model for processing the data trained test. Based on the test result produce the accuracy data is 77.95% and the F1-Score is 57.14%.