Prediction of Indonesian Stock Price Using Combination of CNN and BiLSTM Model

This research investigates the application of deep learning in predicting stock prices, focusing on using a regression model that combines Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (BiLSTM) for the task. Traditional methods of stock price prediction, such as fundamental and technical analysis, exhibit various limitations in handling the stock market's highly volatile and dynamic nature. This study delves into the potential of the CNN-BiLSTM model in effectively capturing and predicting stock price trends, utilizing data collected from daily transactions of the Indonesian Stock Exchange for the last five years The CNN-BiLSTM approach effectively harnesses the strengths of CNN in feature extraction and BiLSTM in recognizing interdependencies within time series data. The primary objective of this research is to develop a model capable of predicting the next day's closing price with improved accuracy. The stages in this study encompass data collection, feature selection, data preprocessing, data splitting, model training, testing, and evaluation. The results indicate that the model achieves superior prediction performance, which utilizes a combination of CNN and BiLSTM architectures with three layers of CNN with 256, 96, and 192 neurons and two layers of BiLSTM with 224 neurons in each layer, optimized with Adam optimizer, yielding an average Mean Absolute Percentage Error (MAPE) of 1.31%, and an average Root Mean Square Error (RMSE) of 92.26.