Predictability Research Of Human Communications

Human being is an existing system, not only has the individuality, but also has a social attribute. Individual behavior has some certain unique characteristics, but social nature has some behaviors such like communicate with surrounding people and transmit information with them. Human behaviors exist in every aspect of our living space, So study and understand human behavior has an important social and economic significance. Communication behavior is an important behavior in all human behaviors. Study statistical characteristics of communication behavior can provide a theoretical basis for further applications. To better understand and characterize the features of human behavior, we analyze the empirical data and find some regular patterns. In turn, we can use these rules and features to serve society better. For example, By means of studying the regularity of communication behavior, we can come to an conclusion that people prefer certain place at a certain time to call certain person, send SMS messages, send mail, etc. So that we can take advantage of these rules to reasonably layout base, reasonably design the communication channel, reasonably layout server, carry out real-time recommendations and conduct business package to achieve precision marketing.Based on380,000individuals’ communication data, this paper mainly study the predictability of communication partners’sequence. Analyze the origin of predictability in human communications and study the medium-range memory length of communication behavior.(1) This paper use information theory methods to characterize the predictability of the communication partners’ sequence. Predictability refers to via the current communication partner to predict next communication partner. In this paper, we show that all the three manners have apparent predictability in partners’patterns, and more over, the short message users’sequences have the highest predictability among the three. We also reveal that people with fewer communication partners have higher predictability.(2) By study the origin of predictability, which comes from two aspects:one is the intrinsic pattern in the partners sequence, that is, people have the preference of communicating with a fixed partner after another fixed one. The other aspect is the burst, which is communicating with the same partner several times in a row. The high burst in short message communication pattern is one of the main reasons for its high predictability, the intrinsic pattern in e-mail partners sequence is the main reason for its predictability, and the predictability of cell-phone call partners sequence comes from both aspects.(3) Based on the communication data and theoretical simulation data, We analyze the inter-event time of human communications and confirm that the behavior of human communication is a combination of two dynamics with different scales. One is the long time scale of Poisson process with no memory, the other is the short time scale of power-law process with memory. The medium-range memory length is defined as the boundary of the power-law and exponential distribution. By defining a threshold value and rescaling the inter-event time, the medium-range memory length has been found. The existence of two time scale dynamics is also verified. Our results can be used to improve the service quality of websites and communication corporations. It also provides a theoretical reference to predict and forecast the natural phenomena.