Identifying Key Nodes for the Influence Spread Using a Machine Learning Approach

Area of Science:

• Network Science
• Machine Learning
• Complex Systems Analysis

Background:

• Identifying key nodes is crucial for applications like viral marketing and epidemic control.
• Machine learning (ML) methods show promise but require refinement for accuracy and generalization.

Purpose of the Study:

• To develop an enhanced ML framework for identifying key nodes in complex networks, specifically for the Independent Cascade model.
• To address challenges in obtaining training labels and improving model generalization.

Main Methods:

• Introduction of a novel "Smart Bins" technique for improved label generation in ML training.
• Development of an ML-based framework to predict influence spread and node characteristics.

Main Results:

• "Smart Bins" demonstrate superiority over existing methods for generating training labels.
• The proposed framework accurately predicts node influence and reveals additional spreading process characteristics.
• Extensive testing confirms the framework's robust generalization across diverse network structures and sizes.

Conclusions:

• The enhanced ML framework offers a significant advancement in identifying key nodes for influence spread prediction.
• The "Smart Bins" method and the framework's ability to extract further spreading characteristics represent novel contributions to network science.