Prediction of Stock Price Data of PT. Ramayana Lestari Sentosa Tbk. using Long Short Term Memory Model
Luki Setiawan, Nurul Fathanah Muntasir, Syafa Fahreza, Asep Sholahuddin · 2021
The investment that Indonesian people are interested in during the pandemic is investing in stocks. However, stock investment will not only generate profits but also risk losses so that to minimize the risk of loss, forecasting is required. One of the prediction methods that can be used is Long Short Term Memory (LSTM) for stock data modeling, this is because LSTM is suitable for predicting and processing important events with relatively long intervals and delays in the time series data. The data used in this research is time series data of PT. Ramayana Lestari Sentosa Tbk. in the period 2 March 2020 to 15 September 2021 with LSTM model using the same number of epochs but different number of nodes, activation functions and optimizers which show varying results and accuracy levels. The selection of the number of epochs 5 is the optimal number of epochs selected. In the experiment, the researchers use 2 types of nodes (4 and 16), 4 types of activation functions (Linear, Sigmoid, Relu, and Tanh), and 3 types of optimizers (Adam, SGD, and Adagrad). The results of the experiment show that the best LSTM model based on the smallest RMSE test value is 15.37 on the use of LSTM model with 4 nodes, Linear activation function, and Adam optimizer. Meanwhile, the stock price prediction for the next day (16 September 2021) obtained based on LSTM model is IDR 639.05. Based on this, it can be concluded that in predicting RALS stock price data using LSTM model with the number of nodes 4 is better than the number of nodes 16. Linear activation function is also better than the activation function of Sigmoid, Relu and Tanh. Then, Adam optimizer is better than SGD and Adagrad optimizer.