Improving IndoBERT for Sentiment Analysis on Indonesian Stock Trader Slang Language

Abstract

Recently, more people access mobile stock trading apps and investors send messages, comments, and posts. Interest in performing sentiment analysis of these messages to predict stock price changes requires ever-improving machine learning models, though, this requires identifying Bahasa Indonesian slang phrases in comments and posts. For developing the model to perform a sentiment analysis on stock price changes, we retrieved data from comments and posts on third-party applications. In the current paper, we presented such a model and test data acquisition using datasets manually labelled by the authors. Our sentiment analysis approach was implemented with a fine-tuned IndoBERT model and achieved 60.35% accuracy predicting the sentiment of 1289 records comments, and posts which better than previous research study. By testing the model, it can do a sentiment analysis on stock price changes and is also capable of identifying the number of slang phrases in the comments and posts by Indonesian traders.