Extended methods for influence maximization in dynamic networks.

Area of Science:

• Social Network Analysis
• Information Diffusion Models
• Computational Social Science

Background:

• Rumor spreading in social networks is analogous to information diffusion.
• The scale of diffusion is highly dependent on the selection of initial nodes.
• Identifying influential nodes is crucial for viral marketing and is known as the influence maximization problem.

Purpose of the Study:

• To develop efficient approximation methods for the influence maximization problem in dynamic networks.
• To extend existing static network methods to handle the complexities of dynamic network structures.

Main Methods:

• Proposed three novel approximation algorithms: Dynamic Degree Discount, Dynamic CI, and Dynamic RIS.
• These methods adapt established techniques from static networks to dynamic network environments.
• Evaluated performance against benchmark methods like MC Greedy and Osawa.

Main Results:

• The proposed methods (Dynamic Degree Discount, Dynamic CI, Dynamic RIS) demonstrated computational times over 10 times faster than MC Greedy.
• Compared to Osawa, the new methods achieved similar performance but were approximately 7.8 times faster.
• The proposed methods are computationally tractable for large-scale dynamic networks.

Conclusions:

• The developed methods are suitable for influence maximization in dynamic networks.
• Further research is needed to determine optimal method selection strategies for specific dynamic networks.
• Parameter tuning for Dynamic CI and Dynamic RIS remains an area for future investigation.