Influence maximisation towards target users and minimal diffusion of information based on information needs

• 主要作者: Temitope, Olanrewaju Abdus-Samad
• 格式: Thesis
• 語言: English; English; English
• 出版: 2020
• 主題: T58.6-58.62 Management information systems
• 在線閱讀: https://etd.uum.edu.my/8398/1/s901087_01.pdf; https://etd.uum.edu.my/8398/2/s901087_02.pdf; https://etd.uum.edu.my/8398/3/s901087%20references.docx; https://etd.uum.edu.my/8398/

Influence maximisation within social network is essential to the modern business. Influence Maximisation Problem (IMP) involves the minimal selection of influencers that leads to maximum contagion while minimizing Diffusion Cost (DC). Previous models of IMP do not consider DC in spreading information towards target users. Furthermore, influencer selection for varying information needs was not considered which leads to influence overlaps and elimination of weak nodes. This study proposes the Information Diffusion towards Target Users (IDTU) algorithm to enhance influencer selection while minimizing the DC. IDTU was developed on greedy approach by using graph sketches to improve the selection of influencers that maximize influence spread to a set of target users. Moreover, the influencer identification based on specific needs was implemented using a General Additive Model on four fundamental centralities. Experimental method was used by employing five social network datasets including Epinions, Wiki-Vote, SlashDot, Facebook and Twitter from Stanford data repository. Evaluation on IDTU was performed against 3 greedy and 6 heuristics benchmark algorithms. IDTU identified all the specified target nodes while lowering the DC by up to 79%. In addition, the influence overlap problem was reduced by lowering up to an average of six times of the seed set size. Results showed that selecting the top influencers using a combination of metrics is effective in minimizing DC and maximizing contagion up to 77% and 32% respectively. The proposed IDTU has been able to maximize information diffusion while minimizing DC. It demonstrates a more balanced and nuanced approach regarding influencer selection. This will be useful for business and social media marketers in leveraging their promotional activities.