Local Community Detection from Noisy Graph Streams
Yang Han, Lingxiang Wang, Qingren Wang, Huabin Wang, Yiwen Zhang, Hong Zhong, Dengcheng Yan · 2024
The demand for real-time graph analysis has encouraged the research of online community detection methods. However, these methods often assume clean data, which is uncommon in reality. How to further improve the performance and robustness of online community detection for real-world networks has garnered the great attention. In this paper, we propose NgsLCD, an algorithm that identifies valid edges in noisy streams to expand communities and mitigate noise effects. It also uses a node-community similarity metric to filter out incorrectly added nodes. Tests on real networks show NgsLCD reduces noise impact and maintains robust detection performance across varying noise rate.