Double dynamic scaling in human communication dynamics

- In this work, Our contribution is to provide further evidence for the view that the initiator of bursty activity is a Poisson process.
- The paper prove again that even if an inter-event time follows power-law distribution which is guaranteed by KS test, the inherent dynamics are double scaling and the special threshold which is called midrange memory length can be calculated.
- Compare with previous research, we introduce a method to mark the random initial event from the complex dynamics.
- The detection of the random initial event has widely application in topic clustering and topic identification.

In the last decades, human behavior has been deeply understanding owing to the huge quantities data of human behavior available for study. The main finding in human dynamics shows that temporal processes consist of high-activity bursty intervals alternating with long low-activity periods. A model, assuming the initiator of bursty follow a Poisson process, is widely used in the modeling of human behavior. Here, we provide further evidence for the hypothesis that different bursty intervals are independent. Furthermore, we introduce a special threshold to quantitatively distinguish the time scales of complex dynamics based on the hypothesis. Our results suggest that human communication behavior is a composite process of double dynamics with midrange memory length. The method for calculating memory length would enhance the performance of many sequence-dependent systems, such as server operation and topic identification.