<![CDATA[Predicting stock prices in the banking sector, particularly for high-capitalisation stocks such as Bank Rakyat Indonesia (BBRI), remains challenging amid market volatility. While Hybrid LSTM-GRU models have demonstrated capability in capturing temporal dependencies in time-series data, prior studies have predominantly focused on manual tuning or optimization of single recurrent architectures, with limited application of Genetic Algorithms for optimizing hybrid recurrent networks in emerging stock markets (R1). This research aims to address this gap by implementing an evolutionary optimization framework using a Genetic Algorithm (GA) to automatically tune the hyperparameters of a Hybrid LSTM-GRU model for enhanced stock price forecasting accuracy. Historical BBRI data from November 2020 to June 2025 were preprocessed through normalization and transformed into supervised time-series sequences before being divided into training, validation, and testing sets. The GA was configured with a population size of 20, 80 generations, and a crossover rate of 0.8 to search for optimal learning rates, batch sizes, and hidden units. The optimized configuration identified 64 units for LSTM and GRU layers, a learning rate of 0.002, and a batch size of 16. The resulting model achieved an RMSE of 82.11 and an MAPE of 1.51%, representing a 20% error reduction compared to baseline hybrid models and outperforming benchmark approaches reported in prior studies (R1). Achieving a 1.51% MAPE indicates reliability for financial forecasting, supporting risk-sensitive investment decision-making (A). Overall, this study demonstrates that evolutionary hyperparameter optimization enhances hybrid deep learning architectures.]]>
