<![CDATA[The stock market is one of the important factors in representing the economic condition of a country. Therefore, a predictive model in analyzing the movement of stock values is needed. This research uses several architectures such as Elman, Long Short-Term Memory (LSTM), and Gate Recurrent Unit (GRU). The problem is how to determine the strategy for tuning parameters. Most likely the strategy can result in a waste of time and resources. This study aims to compare the Genetic Algorithm (GA) and gridsearch in terms of performance and computational time. Elman architecture through GA optimization (Elman-GA) has a Root Mean Squared Error (RMSE) of 165.33 and Elman architecture through gridsearch (Elman-GS) produces 154.47. Elman-GA is much faster with 4874.51 seconds while Elman-GS takes 7148.7 seconds. LSTM architecture through GA optimization (LSTM-GA) has a RMSE of 113.36 while LSTM through gridsearch (LSTM-GS) produces a RMSE of around 111.94. LSTM-GA is also faster because it is only 8733.86 seconds while LSTM-GS takes 16537.42 seconds. The GRU architecture through GA optimization (GRU-GA) has an RMSE of 120.19 and the GRU architecture through gridsearch (GRU-GS) produces an RMSE of 121.35. GRU-GA is much faster because it only takes 6996.62 seconds while GRU-GS takes 19826.86 seconds.]]>
Copyrights © 2022
