Forecasting technology convergence with the spatiotemporal link prediction model

Abstract

Technology convergence represents an innovative process wherein two or more existing technologies amalgamate to form hybrid ones, thereby altering the competitive advantage of organizations and restructuring the competition rules and market networks. Consequently, both researchers and managers are actively engaged in comprehending and forecasting the trend of technology convergence to effectively adapt to and embrace environmental uncertainties. However, existing research on technology convergence primarily focuses on spatially single-dimensional predictions with a relatively short-term horizon of 1–2 years. Additionally, these models often fall short in addressing the issue of imbalanced data within technology convergence networks. In response, we propose the Spatiotemporal Feature Concatenation with Graph Gated Network (STFCGG), a deep learning-based spatiotemporal link prediction model. Our link prediction model achieves simultaneous spatiotemporal predictions, provides medium-to long-term forecasts spanning 3–4 years, and addresses the challenge of imbalanced data from an algorithmic perspective. Experimental results with patent data from the Virtual Reality (VR) and Augment Reality (AR) fields have demonstrated our model's superiority and robustness in handling data imbalance issues, thereby offering valuable insights for future technology convergence directions. In addition to the methodology contribution, our novel link prediction model also provides executives with a valuable tool to develop technological management strategies.