Dynamic community detection as an extension of static community detection, aims to identify similar and closely related sets of nodes across different snapshots and track the evolution of these sets over time, thus revealing the community structure and dynamic changes in dynamic networks. However, the characteristics of high noise and outliers in dynamic networks make traditional Non-negative Matrix Factorization (NMF) unable to maintain robust stability across continuous snapshots. Moreover, for dynamic networks that change complexly over time, it is important to effectively capture and utilize historical information. To address these issues, this paper introduces self-supervised learning. This paper proposes a self-supervised Symmetric Nonnegative Matrix Factorization temporal network community evolution exploration framework (S³ONMF). This model performs temporal coupling between sequences based on multiple random initializations and adaptive weighting, and uses a self-supervised method to screen the dynamic similarity matrix. This effectively resolves the long-standing trade-off between single-snapshot accuracy and temporal consistency. Finally, this Model explores the network evolution patterns between different network snapshots through the local-global evolution pattern (LEP-GEP). Experiments conducted on two types of synthetic dynamic networks and two types of real dynamic networks demonstrate that the proposed framework outperforms state-of-the-art models in dynamic community detection.

Detecting anomalies in dynamic networks is essential for a range of real-world applications, such as social networks and cybersecurity. However, it encounters substantial challenges due to the diverse and ever-evolving nature of these anomalies. We present Attention-based Anomaly Detection In Dynamic Network (AADDN), an innovative end-to-end anomaly detection framework that utilizes attention mechanisms to capture the complex interactions between node attributes (individual characteristics) and structural features (collective patterns) across various time stamps. Unlike traditional methods that depend on heuristic rules with limited scope, AADDN employs a dual autoencoder architecture to learn comprehensive representations in the latent space, allowing the model to more effectively identify both individual and collective anomalies. By emphasizing the integrated learning of temporal, structural, and attribute information, our approach surpasses existing methods, showcasing superior anomaly detection capabilities in dynamic network environments.