Depression Levels Detection Through Twitter Tweets Using RoBERTa Method

Abstract

Mental health conditions are one thing that needs to be considered as important as physical health. Depression is a mental disorder that can affect a person's social life. The Twitter social media platform is where users can pour their hearts out in the form of tweets. This is often the background of a person's level of depression. Accessible and diverse interactions on Twitter considerably influence the psychological condition of its users. This study aims to detect the level of depression through data obtained from Twitter social media tweets. The method used is RoBERTa which is an optimized BERT retraining. Several scenarios are used to get the best accuracy results in text classification. To get the evaluation results, it is necessary to measure using the Confusion Matrix method. The accuracy value is 72% based on numerous tests that have been done.