Weight prediction in complex networks based on neighbor set.

Area of Science:

• Network Science
• Data Science
• Graph Theory

Background:

• Link weights are fundamental to network functionality.
• Real-world network data often suffers from missing information, including link weights.
• Accurate link weight prediction is vital for network analysis and understanding.

Purpose of the Study:

• To develop and validate a novel method for link weight prediction.
• To leverage local network structure (node neighbors) for accurate predictions.
• To assess the method's efficacy in scenarios with missing links and weights, and with missing weights only.

Main Methods:

• A new weight prediction technique is proposed.
• The method utilizes the local network structure, specifically the set of neighbors for each node.
• Performance is evaluated through empirical experiments on real-world networks.

Main Results:

• The proposed method demonstrates accurate link weight predictions.
• Validation was performed in two distinct scenarios: missing links with weights, and missing weights only.
• Experiments confirm the method's effectiveness across different data incompleteness types.

Conclusions:

• The novel method provides reliable link weight predictions based on local network topology.
• The approach is effective even when dealing with significant data loss in network structures.
• This work contributes a valuable tool for analyzing and understanding incomplete weighted networks.