DMSTA-LSTM: A Dual-Stage Multi-Time Scale Temporal Attention-Based LSTM Framework for Forecasting International Financial Indices Across Diverse Varied National Contexts

In the dynamic and complex environment of financial markets, the precision of forecasts is paramount for informed decision-making and risk mitigation. Confronted with the markets’ inherent unpredictability and fluctuations, which frequently precipitate financial disturbances, this study introduces a Dual-Stage Multi-Time Scale Temporal Attention-Based Long Short-Term Memory (DMSTA-LSTM) Framework to refine the predictive accuracy of financial index stocks across diverse countries. Leveraging the LSTM’s capability to capture long-term dependencies, the DMSTA-LSTM model incorporates a pioneering cross-time scale strategy alongside an attention mechanism, effectively amalgamating historical data over 1, 4, 8 and 16 time step spans to markedly enhance multi-step financial forecasting precision. Subjected to rigorous evaluation on the expansive SIHNSF dataset, which encompasses key worldwide stock indices, the DMSTA-LSTM model undergoes comprehensive testing across six real-world stock index datasets at four distinct temporal resolutions. The findings underscore the model’s superior performance, with notable reductions in MSE to 1444.928 (SPX), 14626.177 (IXIC), and 61041.059 (HSI), among others. Additionally, the model achieves exceptionally high R2 values of 0.99705 (SPX) and 0.99819 (IXIC), reflecting its impressive predictive accuracy. The MAE and SMAPE metrics further corroborate the model’s effectiveness, with values such as 30.184 (SPX) and 0.882% (SPX) respectively. By consistently outperforming baseline models across crucial performance indicators, the DMSTA-LSTM model not only transcends contemporary state-of-the-art forecasting methodologies but also sets a new standard in financial market analytics. This development heralds a significant advancement for both investors and analysts, furnishing them with a potent analytical tool for navigating the complexities of financial markets.