FORECASTING THE STOCK PRICE IN VIETNAM: AN APPLICATION OF THE BIDIRECTIONAL LONG SHORT-TERM MEMORY NEURAL NETWORK

Abstract

Accurate stock market prediction can minimize investment risks for investors, improve the efficiency of investment returns, and promote the stable development of the market. However, the high frequency and substantial noise of stock price sequences make precise forecasting a challenging task. Machine learning and deep learning, including enabling computers to perform tasks that typically require human intelligence, are now the dominant trends in stock market forecasting. In this paper, a Bidirectional Long Short-Term Memory (BiLSTM) neural network from deep learning is applied to predict stock price trends in the Vietnamese stock market. The forecasting performance of the one-directional Long Short-Term Memory (LSTM) network and the Recurrent Neural Network (RNN) is compared. Using VN-Index data from 2010 to 2024 and technical indicators, including the Simple Moving Average (SMA), the Moving Average Convergence Divergence (MACD), and the Relative Strength Index (RSI), results show that the BiLSTM model achieves the highest prediction accuracy, nearly 99%. It effectively captures both past and future data information, predicting both short-term and long-term dynamic trends of financial time series. This demonstrates the suitability of the BiLSTM model for stock price forecasting in Vietnam.