Node Attribute-enhanced Community Detection in Complex Networks.

Area of Science:

• Network science
• Data mining
• Machine learning

Background:

• Community detection traditionally relies on node connectivity.
• Node attributes offer rich information but are underutilized in existing methods.
• Current attribute-based methods are limited to specific data types (binary, categorical, numerical).

Purpose of the Study:

• To introduce kNN-enhance, a novel community detection approach using node attribute enhancement.
• To address limitations of existing methods in handling network sparsity and diverse attribute types.
• To improve the accuracy and flexibility of community detection in complex networks.

Main Methods:

• kNN-enhance integrates a k-Nearest Neighbor (kNN) graph of node attributes into the original network.
• This attribute enhancement alleviates sparsity and noise, strengthening community structures.
• Two algorithms, kNN-nearest and kNN-Kmeans, are used to partition the enhanced graph.

Main Results:

• The proposed kNN-enhance approach demonstrated superior performance compared to state-of-the-art algorithms.
• Analyses on synthetic and real-world networks validated the effectiveness of the method.
• The algorithms successfully handled networks with combined binary, categorical, and numerical attributes.

Conclusions:

• kNN-enhance offers a flexible and effective solution for community detection in attribute-rich networks.
• The method enhances network structure by leveraging node attributes via kNN graph construction.
• The approach is scalable and adaptable for analyzing large-scale networks with diverse attribute types.