Research On Modeling Methodology And Application Of Relationship Agent-based Complex Social Networks

Relationship Agent-based Complex Social Networks (ACSN) is a new methodologythat integrates complex network theory and agent-based computer simulation techniques tosocial research,Instead of using a few mathematical equation the whole complex socialsystem model of reductionism, ACSN try to analyzes the behavior which reality individualestablish relations from the perspective of the relationship which reductionism have merelystudied, use relatively simple rules to establish relationship Agent model in the computer, andmany Agents can interact in the computer so that a simulating model of the social network canbe constructed from bottom up. At last the process of micro behaviors emerge and accumulateto the phenomenon of macro can be mapped and explained by using the emerging attribute ofartificial social network.Before designing the model method, we first order out three critical techniques of thecomplex social network modeling: relationship Agent modeling, the interaction networkbetween Agent, the network results emerge; Then from the three viewpoint of technology weinduce the complex network and the corresponding relation between them. In order to verifythe validity of the model we cut a case-requirements of concrete complex social network–thebooks search net of Douban community. Users are reality version of the Agent supplied in themodel, we use users' actual build edge way and the theoretical build edge two-steps way ofnetwork node to establish a more rational Agent and the link weighted evolution model ofAgents' repeatedly accumulation of behavior experience. Based on the above analysis, wegeneralize users' behavior and get the formal models based on relationship Agent andtransform it to Agent Settings of simulation Douban community. We analyze the emergencerelationship from two angles: the individual and the whole. We design the measurement indexof individual fractal structure and the index chain emerged in the social nets and resourcesnets, in order to reflecting the nonlinear relationship between indexes. Finally we put thespecific relationship Agent model into ASNM to run, and got some emerging resultspredicted.The innovation achievements of this model mainly embodied in the following aspects:Firstly, the relationship Agent model, which separates the agent into the objective part,the subjective part and the social part, can make the agent modeling easier and clearer.Secondly, the two-steps built side means based on cluster and optimizing can take use ofpersonal experience of Agent and other Agents' experience; Thirdly, the side evolution style based on Logistic equation makes the Agent' neighbors identify when they are searching,increase the network searchability.Next is the analysis of the emergent results, in our understanding, the result can emergein both macro level and individual level, so we design the measurement index of individualtree fractal structure. What is different is that based on the edge of a reduced, we transformthe original intensity index to a tree shape result, which has produced the trust of the branch,make the foundation of later searchability. For the emergence of the macroscopic level, notonly to build a index, we construct the level which is absolute short network diameter to thenetwork, the cluster and optimizing behavior from community structure to individual, toensure the last emergence of net.Thirdly, we use such a modeling method to analyze the specific Douban community, theprogress it's macro social capital transform from the individual build edge behavior to thenetworks searchability behavior. In sociology,the macro social capital's emergence is mainlymeasured through the questionnaire or even change the measure index, it makes the meaningof how to finally guide the behavior of individual being doubted. The individual originatedbehavior and the macro emerged network in our modeling method is not designed byourselves, the behavior of individual in the experiment can be put directly to the realisticguidance of policy or functional design.As a result, ACSN modeling approach not only supports and verifies the traditionalconclusions, but also manages to deal with tough problems that traditional methods can'tanalyze and even might able to penetrate into untouched areas.