Machine Learning on Dynamic Graphs: A Survey on Applications

TL;DR

This survey expands the scope of machine learning on dynamic graphs by examining lesser-explored domains, beyond traditional networks like transportation and social graphs. It outlines data models (discrete vs. continuous) and six core technique families (temporal embeddings, RNNs, GCNs, DBNs, matrix factorization, DRL), then details six novel applications with architectures such as TransMOT, BGGRU-based air quality prediction, explainable tunneling graphs, MST-GAT for multimodal anomaly detection, gated ST-GCN for bus loads, and graph-based LSTM for AML. The paper highlights how dynamic graph methods deliver superior performance, interpretability, and real-time potential in these domains, illustrating the versatility of graph-based learning for temporal and multimodal data. Overall, it underscores the practical impact of dynamic graph learning and sketches future directions toward real-time and multi-modal analysis across diverse fields.

Abstract

Dynamic graph learning has gained significant attention as it offers a powerful means to model intricate interactions among entities across various real-world and scientific domains. Notably, graphs serve as effective representations for diverse networks such as transportation, brain, social, and internet networks. Furthermore, the rapid advancements in machine learning have expanded the scope of dynamic graph applications beyond the aforementioned domains. In this paper, we present a review of lesser-explored applications of dynamic graph learning. This study revealed the potential of machine learning on dynamic graphs in addressing challenges across diverse domains, including those with limited levels of association with the field.