Local degree blocking model for link prediction in complex networks.

Area of Science:

• Network analysis and mining
• Complex systems science
• Data science

Background:

• Accurate network reconstruction is crucial for understanding complex systems.
• Identifying missing or unreliable links is a key challenge in network analysis.
• Existing link prediction methods often rely on specific network structures or parameters.

Purpose of the Study:

• To propose a novel, parameter-free link prediction method for complex networks.
• To introduce the local blocking (LB) predictor based on local network structures.
• To evaluate the performance of the LB predictor against traditional methods.

Main Methods:

• Studying the statistical features of a node's immediate neighborhood (degree block).
• Developing a parameter-free local blocking (LB) predictor using local link density.
• Conducting experiments on six real-world complex networks.

Main Results:

• The LB predictor demonstrates superior performance compared to traditional local similarity-based methods on most tested networks.
• Experimental results show the LB index effectively detects link formation.
• Correlation analysis reveals LB captures features of both preferential attachment and short-path-based indices.

Conclusions:

• The local blocking (LB) predictor is an effective and parameter-free tool for link prediction in complex networks.
• LB's ability to integrate multiple mechanisms makes it a robust link prediction index.
• This method offers a promising advancement in network reconstruction and analysis.