基于Agent网络化博弈的建模与分析

MODELING AND EXPERIMENTAL ANALYSIS OF NETWORKED EVOLUTIONARY GAME BASED ON AGENT

网络化博弈是一种研究社会网络、生物网络、信息网络及交易网络的重要方式。针对网络化博弈建模方法在网络拓扑结构、博弈模型、建模方式上存在的问题,提出一种基于Agent网络化博弈建模方式。以南京地区汽车品牌价格竞争现象为应用场景,将汽车品牌看作Agent个体,并基于现实网络复杂结构特征,引入"强竞争"和"弱竞争"概念,基于演化博弈理论考察有限理性下汽车品牌价格的竞争行为。实验结果表明,基于Agent网络化博弈模型的范数平均相对误差为2.11%,低于传统模型的4.64%和3.57%,模型精度更高,其中违约惩罚力度对汽车品牌价格竞争影响最为显著,利益分配系数和参与合谋初始意愿次之。

Networked evolutionary game is an important way to study social networks,biological networks,information networks,and the Internet.Aiming at the problems in network topology structure,game model and modeling methods,an agent-based networked evolutionary game modeling method is proposed.In the modeling process,the price competition among automobile brands in Nanjing area was used as an application scenario,and automobile brands were considered as individual agents.Based on the complex structure characteristics of the actual network,the concepts of"strong competition"and"weak competition"were introduced,and evolutionary game theory was used to study the price competition behavior of automobile brands under limited rationality.Finally,the results were obtained from the experiments and the conclusion was drawn:the norm mean relative error of the model proposed by this paper is 2.11%,which is lower than the traditional models of 4.64%and 3.57%,and the model has higher accuracy.Among the factors that affect the price competition among automobile brands,the impact of penalties for breach of contract is most significant,followed by the coefficient of benefit distribution and the initial willingness to participate in collusion.