Stock Price Prediction using BiLSTM with Public Sentiment Factor

Abstract

Stock market is one economic driver. It has roles in growth and development of a country. Stock is an attractive investment due to the huge profit. Many people buy and sell their stock. Stock investors try to choose the good investment company to get profits with small risk. Therefore, stock investors need to be careful and must evaluate a company. With machine learning technology, stock prediction problems can be solved. Deep learning is a subset of machine learning with own network. Deep learning has good performance in managing large amounts of data. This study used stock price history data and public sentiment data on a company. The method used in this research is Bidirectional Long-Short Term Memory (BiLSTM). The features used were closing price and compound score value of the public sentiment. Four scenarios were used in finding the best predictive model. The four scenarios use the same test data with different lengths of training data window. From the modelling, predictions with the model built using BiLSTM resulted in the smallest MSE value of 0.094 and the smallest RMSE value of 0.306.