STWalk: Learning Trajectory Representations in Temporal Graphs

TL;DR

STWalk introduces a novel method for learning node trajectory representations in temporal graphs by combining structural graph properties across time, enabling improved analysis of dynamic network behaviors.

Contribution

The paper presents STWalk, a new framework that captures spatio-temporal node trajectories using random walks at current and past time-steps, with two variants for effective embedding.

Findings

Outperforms baseline methods on real-world datasets

Effective for change point detection in temporal graphs

Trajectory embeddings enable interpretable arithmetic operations

Abstract

Analyzing the temporal behavior of nodes in time-varying graphs is useful for many applications such as targeted advertising, community evolution and outlier detection. In this paper, we present a novel approach, STWalk, for learning trajectory representations of nodes in temporal graphs. The proposed framework makes use of structural properties of graphs at current and previous time-steps to learn effective node trajectory representations. STWalk performs random walks on a graph at a given time step (called space-walk) as well as on graphs from past time-steps (called time-walk) to capture the spatio-temporal behavior of nodes. We propose two variants of STWalk to learn trajectory representations. In one algorithm, we perform space-walk and time-walk as part of a single step. In the other variant, we perform space-walk and time-walk separately and combine the learned representations to get the final trajectory embedding. Extensive experiments on three real-world temporal graph datasets validate the effectiveness of the learned representations when compared to three baseline methods. We also show the goodness of the learned trajectory embeddings for change point detection, as well as demonstrate that arithmetic operations on these trajectory representations yield interesting and interpretable results.