Comparative Algorithms for Stock Price Prediction Based on Market Sentiment Analysis

Stock are a piece of paper that shows the right of the investor (the party who owns the paper) to obtain a share of the prospects or wealth of the organization that issues the security and the various conditions that allow these investors to exercise their rights. However, many investors are still unsure about the risk of investing because they are worried that what they are investing is not as expected. Therefore, a deep learning model is needed that is able to predict fluctuating stock prices. In this study, we compared the traditional LSTM and BiLSTM models to find the strengths and weaknesses of the two models in predicting stock prices. The dataset used in this study includes stock data for companies in the United States (US) listed on stock exchanges such as the NYSE, NASDAQ, and MKT. This dataset includes information about 1344 companies operating in the US. In data with sentiment analysis, the traditional LSTM is superior to BiLSTM with an MAE of 0.10993. Whereas in the dataset without sentiment analysis, BiLSTM is superior to the traditional LSTM with an MAE of 0.10213.