This research project explored the use of artificial intelligence to improve stock price prediction through a hybrid MA–GRU–LSTM model. Using approximately 8,600 hourly data points collected over 3.5 years from a company and its three subsidiaries in Hanoi, the team developed a model combining data decomposition with neural network-based prediction.
The proposed model achieved 1,265 MSE and 4.37% MAPE, outperforming the evaluated GRU and LSTM models in the experiment. The research was presented at the 2026 Japan Design, Idea & Invention Expo in Osaka, Japan, where the project received a Gold Medal.