Dual-Path LSTM-Transformer Architecture for Spatiotemporal Forecasting of Global Digital Divide

The persistent global digital divide, characterized by unequal access to information technology and internet infrastructure, remains a major challenge in digital globalization. This study presents a novel Dual-Path LSTM-Transformer architecture designed to enhance spatiotemporal forecasting of internet penetration trends worldwide. By integrating LSTM's strength in modeling temporal dependencies with the Transformer's self-attention mechanism for capturing long-range spatial interactions, the proposed model effectively addresses the limitations of traditional approaches in handling complex spatiotemporal dynamics. Comparative experiments against baseline LSTM and standard LSTM-Transformer models demonstrate superior prediction accuracy of the Dual-Path model, achieving a Mean Squared Error (MSE) of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{0. 5 3}$</tex>. The results reveal its enhanced capability in identifying regional disparities, including robust digital growth in developed regions and persistent access challenges in rural-urban divide areas. This work advances deep learning architectures for spatiotemporal data analysis and provides a computational framework that can inform targeted strategies to promote digital inclusion globally.