Information fusion-based approach for studying influence on Twitter using belief theory

Area of Science:

• Social Network Analysis
• Computational Social Science
• Information Diffusion

Background:

• Twitter is a key platform for information dissemination, making user influence a critical research area.
• Understanding influential users is vital for effective marketing and political strategies.
• Existing methods may not fully capture the complexity of multi-relational user interactions.

Purpose of the Study:

• To propose a novel approach for multi-level influence assessment on multi-relational networks like Twitter.
• To develop a computational model for assessing user influence degree based on diverse relationship types.
• To validate the model's effectiveness using real-world Twitter data.

Main Methods:

• Modeling Twitter relationships as a multiplex graph with nodes (users) and links (retweets, mentions, replies).
• Employing the conjunctive combination rule from belief functions theory to integrate different relation types.
• Experimenting on large datasets from the European Elections 2014 and CLEF RepLab 2014.

Main Results:

• The proposed model accurately assesses user influence by combining multiple interaction types.
• The method demonstrates flexibility in adapting to various analytical needs.
• Numerical results from belief theory integration are precise and reliable.

Conclusions:

• The developed model offers a robust and flexible framework for influence assessment on social media.
• This approach enhances the understanding of information diffusion dynamics.
• The findings have practical implications for campaign strategies and social science research.