DỰ BÁO GIÁ CỔ PHIẾU BẰNG CÁC MÔ HÌNH HỌC SÂU ĐƯỢC TỐI ƯU BẰNG PSO: XÉT TRƯỜNG HỢP MÃ CỔ PHIẾU VCB

Abstract

Accurate stock price forecasting is a central challenge in financial prediction and risk management. This study focuses on forecasting the daily closing price of VCB stock listed on the Ho Chi Minh City Stock Exchange (HOSE) by applying three popular recurrent neural network (RNN) architectures—Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Bidirectional LSTM (BiLSTM). To enhance model performance and reproducibility, we propose a systematic hyperparameter optimization framework based on the Particle Swarm Optimization (PSO) algorithm. The proposed method automatically searches for optimal configurations of key hyperparameters (number of hidden units, dropout rate, learning rate, and batch size), thereby reducing the need for manual trial-and-error tuning. Experimental results demonstrate that PSO-based optimization significantly improves predictive accuracy across all three models, with the GRU architecture consistently outperforming its counterparts both before and after optimization, achieving RMSE = 0.6851 and R² = 0.9693. The study contributes a practical and replicable optimization approach and provides empirical evidence from the Vietnamese stock market, where comparative deep learning research remains limited. These findings carry important practical implications, enabling investors and portfolio managers to design evidence-based trading and risk management strategies that leverage recent advances in artificial intelligence within the field of financial management.