APPLICATION OF ECONOPHYSICS IN PREDICTING INFOBANK15 INDEX USING THE LONG SHORT-TERM MEMORY METHOD

• Main Author: Fasdhia Daniswara, Dimas
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/77454
• Institution: Institut Teknologi Bandung

Complex system is a system that consists of many parts that interact with each other, resulting in emergent properties that cannot be seen or predicted from the individual components’ characteristics. An example of a complex system is the stock market. The stock market has long been a subject of research for economists, mathematicians, and more recently, physicists. One of the phenomena in the stock market that is studied is the movement of stock prices. In this research, the stock movement will be modeled using the Long-Short Term Memory (LSTM) method, which is a type of machine learning that can process nonlinear data such as stock prices. The objectives of this research are to model the movement of the INFOBANK15 composite index, determine the best hyperparameters for modeling, and provide investment recommendations. In LSTM, there are parameters or hyperparameters used to optimize the model learning. The parameters varied in this study are learning rate and batch size, while the other parameters are kept constant. The results of the modeling will be validated based on the values of Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Correlation Coefficient (R). The research findings indicate that the most optimal model is achieved when the number of neurons is 30, with a learning rate of 0.01 and a batch size of 16, resulting in average RMSE, MAPE, and R values of 14.4413, 0.9524, and 0.9609, respectively, in 10 replications. This suggests that LSTM can effectively learn the movement of INFOBANK15 stocks. The best model is then used to predict the movement of the INFOBANK15 index in the period from January to February 2023. The model performs well during this period, with RMSE, MAPE, and R values of 12.687, 0.08, and 0.913, respectively. The predicted and actual values are used to determine the volatility of INFOBANK15, and it is found that the volatility of INFOBANK15 is higher than the predicted volatility, LQ45 index, and IHSG index. Therefore, INFOBANK15 index can be a choice of investment for someone with a high-risk profile. Based on the model’s prediction results, from the end of August 2023 to early December 2023, the INFOBANK15 index experienced a downtrend from mid-August to early December, followed by a rapid recovery towards mid-December.