PSO-OPTIMIZED DEEP LEARNING MODELS FOR STOCK PRICE FORECASTING: EVIDENCE FROM VCB STOCK

Abstract

Accurate stock price forecasting plays a crucial role in investment analysis and financial risk management. This study aims to improve the forecasting accuracy of the daily closing price of VCB stock listed on the Ho Chi Minh City Stock Exchange (HOSE) using deep learning models. Three architectures, namely LSTM, GRU, and BiLSTM, were developed and compared within a unified experimental framework. The forecasting task was formulated as a univariate time-series problem using only the closing price series, with a 10-to-1 sliding window approach in which the prices of the previous 10 trading sessions were used to predict the closing price of the following session. To reduce reliance on manual trial-and-error procedures, Particle Swarm Optimization (PSO) was employed to optimize key hyperparameters, including the number of hidden units, dropout rate, learning rate, and batch size. Experimental results show that PSO significantly improved the forecasting performance of all three models, with the optimized GRU achieving the best results on the test set, with an RMSE of 0.6851 and an R² of 0.9693. These findings provide empirical evidence for the effectiveness of integrating deep learning with hyperparameter optimization in stock price forecasting for the Vietnamese stock market.