Current GNN-based asset price prediction models often focus on a fixed group of assets and their static relationships, overlooking the reality that asset pools and their interrelationships evolve over time.
We propose DySTAGE (ICAIF 2024), a framework with a universal formulation that transforms asset pricing time series into dynamic graphs, accommodating asset addition, deletion, and changes in correlations. Assets at various historical time steps are structured as a sequence of dynamic graphs, where connections reflect long-term correlations. The Topological Module deploys Asset Influence Attention along with Asset-wise Importance Encoding, Pair-wise Spatial Encoding, and Edge-wise Correlation Encoding, while the Temporal Module captures node representations across time via attention. Experiments on three real-world stock pricing datasets show DySTAGE surpasses popular benchmarks in return prediction and offers profitable investment strategies.