Research On The Dynamics Of Multiperson Communication Behavior In QQ Groups

The rapid development of the network economy has greatly enriched people’s daily communication methods.Meanwhile,influenced by epidemic prevention and control,people’s work,study and life rely more on online communication,such as QQ,We Chat,Ding Talk,Tencent Meeting,etc.The interactivity in social networks has been further enhanced,and rich human interaction data has been generated,which helps us to better study the impact of multi-individual interactions in human dynamics.At present,there are few studies on frequent interaction behaviors of groups in the short term,and our research results are helpful to deepen the understanding of human behavior patterns and provide implications for human dynamics modeling from the perspective of multi-individual interaction.It helps to provide optimization suggestions for social software such as QQ,We Chat,Ding Talk,and Weibo regarding application services such as interactive recommendation and precise marketing;and in maintaining network security,information sharing,traffic distribution,etc.It is also useful for maintaining network security,information sharing and flow assignment.Based on 252 QQ groups,we empirically analyzed and modeled the temporal characteristics of human group communication behavior,and the details of our work are as follows:(1)Empirical analysis of human behavioral characteristics in QQ groups.Based on nearly 10 million message sending data of 252 QQ groups,we statistically analyzed the characteristics of group online frequent interaction behavior,and the empirical results show that: different QQ groups show different temporal characteristics,and the distribution of adjacent message sending time intervals shows three different patterns with the increase of group chatting activity,i.e.,bimodal distribution,double-power-law distribution and single-power-law distribution,which is different from the SMS interaction system and the Wikipedia entry editing system,where only a single distribution pattern appears.For the first time,a transition between bimodal,double-power-law,and single-power-law distributions of event occurrence times was found in the same system.Although previous models could also show these three types of distributions,they were presented in different models,respectively.Furthermore,the single-power-law distribution of human collective behavior and the transition between different distributions are only theoretical predictions,and there is no empirical evidence from previous studies,which is verified by empirical data in this paper.Group communication behavior is affected by people’s physiological patterns,sleep mechanisms,and weekend culture,and there are obvious fluctuations and cycles.The activity of all three distributions of groups is negatively correlated with the second-order moments of the time distribution.(2)Mathematical model construction for QQ group communication dynamics.In this paper,we proposed a multi-individual interaction queueing model that simulates the process of sending messages clearly by different members of the group.The model matches well with empirical results in terms of the presentation of three distributions,the transition between different patterns,and the deviation from the empirical results of the standard power law or exponential distribution,and the model is valid and robust.The main dynamical mechanisms explaining similar group online communication behavior are:human decision-driven task queuing processes,individual behavior with Poisson properties,group interactions,and group activation.In addition,the simulation results show that the single-power-law distribution is caused by a combination of periodic elements and high group activity,which is somewhat different from the causes of population growth in the the Wikipedia entry editing system.The degree of group activity and task initiation rate on time dependence determines the temporal distribution pattern of group communication behavior,and the group response has almost no effect on the type of distribution pattern.(3)A study of the influence of central nodes in group chatting behavior.The characteristics of sending message behavior of different individuals in the group were studied from the individual perspective,and the empirical statistics show that the probability distributions of activity,response rate and initiation rate among individuals have power-law characteristics,and there are generally a few individuals who actively send messages and most individuals who rarely participate in communication in different QQ groups.There are positive correlations between the initiation rate and response rate of individuals and individual activity,respectively,but the correlation between initiation rate and response rate is not obvious.It indicates that the behavior of sending messages is complex.Active individuals in the group may result from frequent initiation of topics or active response to others.Then we constructed a multi-individual interaction queue model with heterogeneity in activity among individuals,the results show that the distribution pattern of the system is unchanged in the absence and presence of heterogeneity in individual activity.It indicates that for the emergence of non-Poisson features at the group level,the heterogeneous distribution of activity among individuals is not the underlying cause.Finally,the influence of central nodes on group chat behavior is investigated,and the simulation results show that central nodes with high response rate are more influential than central nodes with high initiation rate.