The Community for Technology Leaders
2015 IEEE First International Conference on Big Data Computing Service and Applications (BigDataService) (2015)
Redwood City, CA, USA
March 30, 2015 to April 2, 2015
ISBN: 978-1-4799-8128-1
pp: 478-486
ABSTRACT
The study of social influence in online social networks has attracted great interests in recent years for its applications in information propagation and marketing. While many existing studies focus on the measurement of social influence on various platforms, there is a lack of comprehensive analysis regarding the effectiveness of the principal features for measuring user influence. In this paper, we employ Entropy method and Rank Correlation Analysis to identify the key manifest features for measuring user influence. We also utilize Principal Component Analysis and Stepwise Multiple Linear Regression to analyze the most important hidden social attributes for identifying influential users on Twitter. Our study reveals a number of novel findings as follows: (i) Firstly, besides mention and rewet actions that have been widely used to measure user influence in literature, we find that number of public lists, new tweets, follower to friends ratio are also fairly effective indicators for user influence, (ii) We further discover that popularity, engagement and authority are the three most important social attributes to drive user influence in Twitter environment, (iii) Finally, we compare four popular influence scoring services, and find that new mentions and number of public lists are the two most effective manifest features for their influence ranking, and popularity is commonly considered as the first key social attribute of the influencers on Twitter.
INDEX TERMS
Twitter, Entropy, Correlation, Media, Measurement, Australia
CITATION

Y. Mei, Y. Zhong and J. Yang, "Finding and Analyzing Principal Features for Measuring User Influence on Twitter," 2015 IEEE First International Conference on Big Data Computing Service and Applications (BigDataService)(BIGDATASERVICE), Redwood City, CA, USA, 2015, pp. 478-486.
doi:10.1109/BigDataService.2015.36
88 ms
(Ver 3.3 (11022016))