A Dynamic Community Detection Method for Complex Networks Based on Deep Self-Coding Network

Area of Science:

• Computer Science
• Network Science
• Artificial Intelligence

Background:

• Community detection is crucial for understanding complex dynamic networks.
• Existing methods struggle with the evolving nature of network structures and node features.
• Dynamic networks present unique challenges for traditional graph analysis techniques.

Purpose of the Study:

• To propose a novel dynamic community detection method for complex dynamic networks.
• To leverage graph convolution neural networks (GCNNs) for improved network analysis.
• To enhance the accuracy and robustness of community detection in evolving networks.

Main Methods:

• A dynamic community detection method based on GCNNs was developed.
• An encoding-decoding mechanism was designed for node feature reconstruction.
• Multiple graph convolutional layers formed the encoder, and a two-layer perceptron served as the decoder.
• Stochastic gradient descent was employed for optimization.

Main Results:

• The proposed model demonstrated improved detection accuracy on benchmark datasets (Karate Club and Football).
• The Normalized Mutual Information (NMI) metric saw an average improvement of 7.65%.
• The method effectively mitigated the issue of node oversmoothing in dynamic networks.

Conclusions:

• The GCNN-based dynamic community detection method offers superior performance.
• The approach provides a robust solution for analyzing complex and evolving network structures.
• This method advances the field of network science and community detection algorithms.