An adaptive heuristic clustering algorithm for influence maximization in complex networks.

Area of Science:

• Network Science
• Computer Science
• Data Analysis

Background:

• Influence maximization is crucial for understanding and controlling spread in networks (e.g., diseases, innovations, viruses).
• Existing methods like Katz index and LP index have limitations in complexity or application scope.
• Efficiently identifying influential nodes is key for targeted interventions and strategic seeding.

Purpose of the Study:

• To develop a more effective and efficient influence maximization strategy.
• To introduce a novel node similarity measure and a clustering algorithm for seed set selection.
• To balance precision and computational complexity in influence maximization.

Main Methods:

• Proposed a dynamic local similarity index, a path-based node similarity measure adaptable to network topology.
• Developed a novel strategy for selecting initial cluster centers using extended neighborhood coreness and minimum distance.
• Introduced an adaptive heuristic clustering algorithm to find seed sets with maximum collective influence.

Main Results:

• The dynamic local similarity index offers a better balance between complexity and precision compared to existing indices.
• The clustering strategy accelerates convergence and avoids local optima, particularly in non-connected networks.
• Empirical results on four real datasets demonstrate the proposed algorithm's superior effectiveness and efficiency against state-of-the-art methods.

Conclusions:

• The proposed dynamic local similarity index and adaptive clustering algorithm provide a robust solution for influence maximization.
• These advancements can enhance the analysis, prediction, and control of spreading processes in diverse real-world networks.
• The findings offer practical improvements for applications ranging from marketing to epidemiology.