Utilization of LSTM (Long Short Term Memory) Based Sentiment Analysis for Stock Price Prediction

Abstract

This study aims to utilize sentiment analysis in predicting stock price movements. Sentiment analysis can provide information to investors to understand market sentiment. This study uses a text-based approach by pre-processing data, constructing a sentiment analysis model and evaluating model performance. The collected data is analyzed to identify the text's positive, negative, or neutral sentiments. The approach used in scoring sentiment analysis is the Text blob approach and the Lexicon approach. Differences in the results of the accuracy of the two Sentiment Analysis approaches with the LSTM model have an influence on the prediction results with a better increase in accuracy using the Lexicon Sentiment Analysis approach. Then the LSTM model is implemented to classify texts into the desired sentiment categories. The results of this study are insight into the use of sentiment analysis in predicting stock price movements. The implemented sentiment analysis model can be a useful predictive tool for investors and stock practitioners in making investment decisions.