Research On Empirical Study And Mechanisms Of Human Dynamics

With the advances in technology and development of the Internet, great changes occur in the way of communications, from the initial letters, e-mail, instant messaging, blog in Web1.0era to microblogging in Web2.0era. Recently, microblogging is becoming an essential communication tool and plays an increasingly important role in the field of information dissemination. The dynamics of many technological, social phenomena are influenced by human behavior, turning the understanding of human dynamics into a central question of modern science. Current models of human dynamics assume that human behavior can be well approximated by Poisson Processes. In contrast, there is increasing evidence that many human actions follow non-Poisson statistics, characterized by bursts of rapidly occurring events separated by long periods of inactivity. Many researchers focused on the mechanisms of the bursty nature, the most famous one is the Barabasi model based on a decision-based queuing process, in which the individuals execute tasks based on some perceived priority. Meanwhile, other researchers also proposed various generating mechanisms based on interest, adaptive adjustment, memory and selective theory and successfully made progress in some respects. Along with the development of microblogging, research on its statistic characteristics is becoming more and more essential. Meanwhile, the mechanism of human dynamics is still an open question and need more researchers pay more attentions in it. This study is conducted based on the above age and academic background.On basis of determining the research objects, we start the research on sequences of empirical research, simulation and improvement of Barabasi Model and other generating mechanisms. In the aspect of empirical research, we analyze the microblogging users behavioral data and make statistics based on whether the inter-event time follows power law distribution or exponential distribution. In the aspect of simulation and improvement of Barabasi model, we analyze relationships between the power law index and arrival rate, service rate and priority function. Meanwhile, also with simulation method, we extend the Barabasi model by changing the arrival distribution, service distribution and priority function and get some valuable conclusions. In the end, we select an M/D/1priority queuing system with vacations and interest-driven model to model the behavior of microblogging users. In addition, we propose an innovative model of human dynamics based on diminishing marginal utility. Through the theoretical proof, we derive the effectiveness of those models on explaining the non-Poissonian characteristic.Under the guidance of the research aims, we focus on our research and the main contents of this research are as followings.Firstly, starting from the empirical data-the microblogging users behavioral data, we analyze whether the behavior that users publish microblog follow non-Poisson statistics based on whether the inter-event time follows power law distribution or exponential distribution. Meanwhile, we also analyze other indicators such as the burstiness, memory and relationship between activity and exponential index.Secondly, we analyze the most famous mechanisms of non Poisson statistics-Barabasi model, which is based on a queuing process with priority. The research can be divided into three aspects. In the first place, with discrete events model, we simulate Barabasi model with MATLAB software and obtain the waiting time distribution. In the second place, we analyze relationships between the power law index and arrival rate, service rate and priority function. Lastly, also with simulation method, we extend the Barabasi model by changing the arrival distribution, service distribution and priority function and get some valuable conclusions.Thirdly, based on deeply analyzing the Barabasi model, we propose the limitation of its application. With the combination of microblogging characteristics, we select an M/D/1priority queuing system with vacations to model the behavior of microblogging users. By compare the theoretical results with the empirical data, we prove that the effectiveness of this queuing model with vacations.Fourthly, we focus on other generating mechanisms besides the Barabasi model, including two aspects. In the first place, we modify the interest-driven model by changing the interest function and challenge this interest-driven model. In addition, based on the nature of human selection and internal driving force, we propose an innovative model of human dynamics based on diminishing marginal utility.This article is a comprehensive study of the classic literature of human dynamics, and we also make some valuable innovations in the empirical study and generating mechanisms based on the former research. The innovations are as following:Firstly, the empirical analysis of human dynamics of microblogging data is not involved in the existing literature, so this empirical research of microblogging flourish the empirical study of human dynamics. By analyzing the current development status of microblogging, we determine the user behavioral data of Sina Microblogging as our study object. With the crawler software, we Crawl the behavioral data of the top one hundred famous users based on their followers and then derive that the interval distribution between publishing follows a power law distribution with the average power law index is1.34, which does not belong to the1or1.5class. And we also find that the publishing behavior has strong bursty and weak memory, whose average of burstiness is0.25and memory is approximately zero. Meanwhile, we also derive that the power law index has a strong positive correlation with users’activity degree. The index will increase until the limitation to1.5along with the growth of users’activity.Secondly, we analyzing the factors which will influence the power law index which is also not appear in former study. Besides, we improve the Barabasi model and get large index range, which is more suitable to the empirical data. And this research is also new built. In this paper, we analyze the most famous mechanisms of non Poisson statistics-Barabasi model, which is based on a queuing process with priority. The research can be divided into three aspects. In the first place, with discrete events model, we simulate Barabasi model with MATLAB software and obtain the waiting time distribution. After testifying the simulation results with empirical data, we show that our method to simulate the Barabasi model is effective. In the second place, we analyze relationships between the power law index and arrival rate, service rate and priority function and get the following conclusions:the power law trend is becoming weak with the decreasing of arrival rate and when the rate comes to zero, the interval distribution follows exponential distribution. And along with the decreasing of arrival rate, the power law index also decreases until the limitation to1. And we get the finding that the index is not only just1or1.5, which ranges between1and1.6. Also along with the service rate, the interval distribution does not changed greatly, but the index also decreases with the service rate until limitation to1.5. Lastly, also with simulation method, we extend the Barabasi model by changing the arrival distribution, service distribution and priority function and get some valuable conclusions. When the arrival process follows power law distribution with index1, the interval distribution still follows power law with large index range between1and5. However, if the service process follows power law, the interval distribution is also power law and the index is also still1.5. The priority function has no effect on the interval distribution.Thirdly, based on the characteristics of microblogging, we propose the queuing system with vacations and get the theoretical results by using stochastic point processes. Based on deeply analyzing the Barabasi model, we propose the limitation of its application. With the combination of microblogging characteristics, we select an M/D/1priority queuing system with vacations to model the behavior of microblogging users. With the stochastic decomposition method, we derive the waiting time distribution and get the power law index, which is suitable with the empirical data and prove that the effectiveness of this queuing model with vacations.Lastly, we do some research on mechanisms based on interest and marginal utility and derive the theoretical results. According to people’s real life experience, we sum up that people’s interest in certain things has a trend that the interest firstly increases and after some time it decreases. Then we modify the interest-driven model by changing the interest function and get a more realistic model of human dynamics. With the stochastic point process theory, we analyze this model and get the power-law distribution waiting time. In addition, based on the nature of human selection and internal driving force, we propose an innovative model of human dynamics based on diminishing marginal utility, deriving the power-law distribution both in general utility function and the classic Cobb-Douglas utility function and wide range of power-law indexes.The achievements in the paper provides empirical evidence of non-Poisson statistics, expands the classic Barabasi model, proposes various mechanisms based on queuing system with vacations, diminishing marginal utility theory and modifies the interest-driven model. This paper has significant roles to understand the non-Poisson characteristic, the generating mechanisms and the dynamical effectiveness.