Uncovering multi-scale characteristics of users in microblogsphere

Extremely popular and commonly adopted microblogging service (e.g., Twitter) has emerged as a new form of communication. The unprecedent population and open nature of this publishing venue opens a vast opportunities for individuals and organizations to popularize an idea or market a product well beyond his/her social network. This is followed by diverse users behavior patterns corresponding with different targets. Moreover, these human actions are regulated by the mechanism that operates across multiple scales. In this paper, we investigate users´ characteristics in microblogsphere, from the perspectives of probability distribution and timing correlation. The former viewpoint reveals scaling law of inter-publishing timing distribution, “flat head” phenomenon and “hump” in the log-log slope. We further examine the timing correlation by using multi-scale entropy analysis, aiming to get further insights into behavioral patterns and build a new constraint on measuring the precision of a behavioral model. In addition, a type of users with constantly periodic pattern is distinguished by using multi-scale entropy analysis.