A Supervised Link Prediction Method Using Optimized Vertex Collocation Profile.

Area of Science:

• Network Science
• Data Mining
• Graph Theory

Background:

• Traditional link prediction methods struggle with vertex information accessibility and heuristic topological analysis.
• Network embedding models offer efficiency but lack interpretability in link prediction tasks.

Purpose of the Study:

• To propose a novel, interpretable link prediction method addressing limitations of existing approaches.
• To enhance the representation of topology context using subgraph information.
• To reduce computational complexity through community detection.

Main Methods:

• Introduced a 7-subgraph topology to capture vertex context.
• Developed an optimized vertex collocation profile (OVCP) to generate unique, interpretable vertex feature vectors.
• Integrated OVCP features with a classification model and employed overlapping community detection.

Main Results:

• The proposed OVCP method achieved promising performance in link prediction.
• Demonstrated superior interpretability compared to network-embedding-based methods.
• Overlapping community detection significantly reduced computational complexity.

Conclusions:

• The OVCP method offers an effective and interpretable solution for link prediction.
• This approach advances the field by combining detailed topology context with explainable features.
• The method shows potential for application in complex real-world networks.