Influence Maximization for Fixed Heterogeneous Thresholds

Area of Science:

• Network Science
• Computer Science
• Data Science

Background:

• Influence maximization is crucial for targeted information diffusion in networks.
• The problem is computationally challenging (NP-hard), requiring efficient solutions.
• Existing methods may not capture diverse influencer characteristics effectively.

Purpose of the Study:

• To propose and evaluate two novel metrics for influence maximization: Balanced Index (BI) and Group Performance Index (GPI).
• To assess the performance of BI and GPI under the Linear Threshold Model.
• To investigate the influence of network structure (degree assortativity) and threshold distributions on cascade size.

Main Methods:

• Developed the Balanced Index (BI) by combining node degree, susceptibility, and neighborhood impact.
• Developed the Group Performance Index (GPI) evaluating initiator performance within random sets.
• Tested both metrics on the Linear Threshold Model with fixed, known thresholds.
• Analyzed the effect of degree assortativity and threshold distributions on cascade spread.

Main Results:

• The Balanced Index (BI) efficiently identifies influencers with high resistance and out-degree.
• The Group Performance Index (GPI) effectively measures initiator performance in group settings.
• Both BI and GPI demonstrated strong performance in influence maximization tasks.
• Network degree assortativity and threshold distributions significantly impact cascade size.

Conclusions:

• The proposed BI and GPI metrics offer effective and computationally efficient approaches to influence maximization.
• These metrics provide valuable tools for understanding and optimizing information diffusion in complex networks.
• Further research can explore these metrics across various cascade models and network types.