GAPNet: Plug-in Jointly Learning Task-Specific Graph for Dynamic Stock Relation

TL;DR

GAPNet tackles the misalignment and limited generalization of predefined stock relation graphs in financial forecasting by introducing an end-to-end plug-in that jointly learns task-specific graph topology and representations. It combines a Spatial Perception Layer to expand the node receptive field and a Temporal Perception Layer to capture long-term dependencies, producing a task-aligned adjacency $adj^t$ that can feed any GNN backbone (pairwise or hypergraph). Empirical results on NASDAQ and NYSE show GAPNet-aligned models achieve higher annualized returns and Sharpe ratios across multiple backbones, with robust ablations confirming the value of both SPL and TPL and the approach’s generalizability. The work demonstrates that jointly learning graph structure and representations is crucial for task-specific relational modeling in finance and offers a broadly applicable paradigm for end-to-end graph-based forecasting tasks.

Abstract

The advent of the web has led to a paradigm shift in the financial relations, with the real-time dissemination of news, social discourse, and financial filings contributing significantly to the reshaping of financial forecasting. The existing methods rely on establishing relations a priori, i.e. predefining graphs to capture inter-stock relationships. However, the stock-related web signals are characterised by high levels of noise, asynchrony, and challenging to obtain, resulting in poor generalisability and non-alignment between the predefined graphs and the downstream tasks. To address this, we propose GAPNet, a Graph Adaptation Plug-in Network that jointly learns task-specific topology and representations in an end-to-end manner. GAPNet attaches to existing pairwise graph or hypergraph backbone models, enabling the dynamic adaptation and rewiring of edge topologies via two complementary components: a Spatial Perception Layer that captures short-term co-movements across assets, and a Temporal Perception Layer that maintains long-term dependency under distribution shift. Across two real-world stock datasets, GAPNet has been shown to consistently enhance the profitability and stability in comparision to the state-of-the-art models, yielding annualised cumulative returns of up to 0.47 for RT-GCN and 0.63 for CI-STHPAN, with peak Sharpe Ratio of 2.20 and 2.12 respectively. The plug-and-play design of GAPNet ensures its broad applicability to diverse GNN-based architectures. Our results underscore that jointly learning graph structures and representations is essential for task-specific relational modeling.

Figures (4)

• Figure 1: Training paradigm comparison from $t0$ to $tn$. (a) Two-step training paradigm using static graphs; (b) Two-step training paradigm using rule-based dynamic graphs; (c) End-to-end training paradigm using GAPNet.
• Figure 2: Model architecture of GAPNet and an overview of the end-to-end training paradigm on the stock ranking task.
• Figure 3: Backtesting (top-5) results on NASDAQ dataset with/without GAPNet.
• Figure 4: The comparsion of computational cost for six backbone models. Left: training time; Right: memory usage.