Refining Stock Market Projections: A Comparative Analysis of PSO-Optimized LSTM Architectures and LSTM-GRU Models for Improved Forecast Accuracy

This Global financial market unpredictability demands sophisticated forecasting tools. Conventional approaches, including SVM and LR, have limitations in capturing complex market patterns. We assess the viability of advanced deep learning architectures— specifically Long Short- Term Memory -LSTM networks, LSTM-GRU hybrids, and PSO- optimized LSTM models—to more accurately model these intricate patterns. We assess the impact of PSO Technique as a tuning mechanism to enhance LSTM performance. A comparative evaluation using RMSE and MAPE reveals that PSO-optimized LSTM models significantly outperform standalone LSTM and LSTM-GRU architectures in terms of forecast accuracy.