Explaining Temporal Graph Predictions With Shapley Values

2026-04-27

AI Summary

Existing temporal graph neural networks (TGNNs) suffer from a lack of interpretability, making it difficult to understand how they leverage spatiotemporal information for prediction. This work proposes two model-agnostic, local explanation methods: an event-level explainer based on Shapley values and a feature-level explainer grounded in Owen values, uniquely integrating both to capture the hierarchical dependencies between events and their constituent features. The approach efficiently estimates event contributions via KernelSHAP and performs feature attribution using Owen values. Experimental results demonstrate that the proposed method outperforms current explanation techniques across multiple datasets and evaluation metrics. Furthermore, it successfully identifies the root cause of performance degradation in the TGAT model under sparse explanation scenarios—namely, a flaw in its timestamp extraction mechanism.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Temporal Graph Neural Networks (TGNNs) have become increasingly popular in recent years due to their superior predictive performance by combining both spatial and temporal information. However, how these models utilize the information to make predictions is rather unexplored, leading to potentially faulty or biased models. This work introduces two novel model-agnostic explainers for local explanations of TGNNs based on Shapley and Owen values. The first method, an event-level (edge-level) Shapley explainer, applies the KernelSHAP algorithm to estimate contribution scores for individual temporal events, providing interpretable descriptions for model behavior. The second, a feature-level Shapley explainer, extends this framework by decomposing event-level Shapley values into Owen values, and thereby uncovers hierarchical dependencies of the event and its features. The explainers outperform SOTA explainers on different metrics and datasets. Additionally, the Feature Explainer reveals a faulty extraction of actual timestamps of a commonly used TGAT implementation, helping to further understand performance drops on very sparse explanations.

Problem

• Temporal Graph Neural Networks
• Explainability
• Shapley Values
• Model Interpretability
• Temporal Events

Innovation

• Temporal Graph Neural Networks
• Shapley Values
• Owen Values
• Model Interpretability
• KernelSHAP