Stock Sentiment Prediction of LQ-45 Based on News Articles Using LSTM

The growth in the number of investors in the financial market indicates that the investment world is currently experiencing rapid development. One of the long-term investment instruments that has experienced significant growth in the financial market is the stock market. Growth data as of September 2024 sourced from the Indonesia Stock Exchange report reveals that the number of stock market investors has reached more than 6 million single investor identification (SID). The share price of a company can be influenced by two main factors, namely internal factors and external factors. Internal factors come from within the company itself, while external factors come from conditions outside the company. Model development uses the Long Short-Term Memory (LSTM) method to predict daily stock sentiment in realtime. Labeling is done based on the history of stock price changes taken from Yahoo Finance. Stock market news data is obtained automatically every day through Really Simple Syndication (RSS) with the help of cronjob. The results of the LSTM model showed good performance, with a macro F1-Score of 0.73, a macro precision of 0.72, and a macro recall of 0.75. When compared to baseline models such as Logistic Regression, Naive Bayes, and Random Forest which only achieve a macro F1-Score of 0.58, 0.54, and 0.65, respectively, it can be concluded that the developed LSTM model has superior performance. This research can provide new considerations to investors, so as to reduce the risk of loss due to errors in choosing companies to invest in.