Stock Price Prediction Modeling Using Recurrent Neural Network and Long-Short Term Memory

Stock price fluctuations were very difficult to predict the direction of the changes. There were generally estimated to follow three analysis techniques: technical, fundamental, and sentiment. Technical analysis involves observing prices in the past, fundamental analysis is related to the analysis of ongoing business situations, while sentiment analysis includes stock prices that were affected by business aspects, current information, and business activities. Valid price data of the BCA Company used was the stock price from 2019 until 2024. The purpose of this study is to find alternative models of the RNN and LSTM models. The methods used in this study are the documentation method and the optimization method. Accuracy measurements used Mean Square Error (MSE) and Mean Absolute Error (MAE) metrics. The results of stock prediction using the RNN model got poor results with epoch 10 obtaining an accuracy of 61.3%, while using the LSTM model obtained quite good results with epoch 10 obtaining an accuracy of 87.7%. Stock predictions using the combined RNN-LSTM models were able to get good results with epoch 10 obtaining an accuracy of 93.3%.