Empirical Analysis on the Human Dynamics of a Large-Scale Short Message Communication System

Research on human behavior has attracted increasing attention recently because of its scientific significance and potential applications. Some empirical results have indicated that the timing of many human activities follows non-Poisson statistics. We analyze a real-life huge dataset of short message communication with 6326713 users and 37577781 records during the 2006 Chinese New Year. The results show that the number of short message sendings, the interevent time between two consecutive short message sendings and the response time all follow heavy-tailed distribution. We further observe a strongly positive correlation between the activity and the power-law exponent of the interevent time distribution. In addition, the short message communication system displays a bursty property yet no memory effects, which is in particular different from some well-studied human-activited systems such as email-sending, library-loaning and file printing.