An Efficient Vertex-Driven Temporal Graph Model and Subgraph Clustering Method

Hanlin Zhang

doi:10.1109/ACCESS.2022.3208360

Return

An Efficient Vertex-Driven Temporal Graph Model and Subgraph Clustering Method

2022-01-01

PRE

DOI：10.1109/ACCESS.2022.3208360

Original

Original request for help

Save

Abstract

En 中文

The temporal graph can represent a temporal relationship widely used in compound synthesis analysis, biological gene analysis, etc. However, the temporal graph would embody vertex updates frequently, high time resolution, and not enumerated rules. The construction and update of some temporal graph models are too dependent on the graph operation sequence, which leads to a lack of an effective model. Simultaneously, the temporal subgraph clustering of the temporal graph with frequent updating for the lack of an effective model leads to low accuracy. Therefore, we propose an efficient and frequently updated temporal graph model as vertex driven and corresponding temporal subgraph clustering method. First, we propose a temporal graph construction algorithm and set two thresholds to divide the temporal graph on a timeline to obtain temporal subgraphs. Next, an enhancement strategy based on the sliding window is proposed to accelerate the construction process. Third, we offer a double-standard temporal subgraph clustering method based on community comparison and temporal distance. The temporal subgraph can be effectively distinguished in temporal and structure dimensions. Lastly, experimental results on both real and synthetic datasets show that the temporal graph model proposed in this work can reduce the time overhead of construction compared to other existing models. The cluster method improves the clustering accuracy of temporal subgraphs. The clustering results show through the hierarchical clustering at the same time.

Keywords:

Sensors

Analytical models

Image edge detection

Clustering methods

Data models

Clustering algorithms

Graph modeling

Temporal Graph

temporal graph model

subgraph clustering

sliding window