摘要:
基于复杂网络视角,从微观层面对新产品扩散问题进行研究.结合心理学、社会学和博弈论,构建个体决策行为模型,并设计出一种智能学习方法实现消费者间的互动.在Anylogic上建立多 Agent模型,考虑局部网络效应,设计网络演化规则,探究不同网络环境下网络结构和消费者决策共同演化规律,并分析产品效用参数、沟通交互强度、种子用户数量以及链接断开时间对共同演化的影响.研究发现:①尽管初始网络结构不同,消费者关系结构均能够以更高的聚集性和稳定性重构,且在小世界和无标度网络下,共同演化、彼此促进,从而有助于产品扩散;②尽管在高度数的无标度网络下共同演化会使扩散结果呈现出两面性,但在扩散过程中网络结构变化趋势一致;③网络效应强度提高能增大产品效用值,但过高和过低的效用值都将削弱网络演化对扩散的促进效果;④ 种子用户在达到一定规模后,不再影响最终扩散结果;⑤若共同演化能在最终时刻前达到稳定,链接断开时间则不影响扩散结果.%Based on the perspective of complex network,the problem of new product diffusion is discussed on a microscopic level.First,the customer decision-making behavior model is designed combined with psychology,sociology,game theory and other relative theories.An intelligent learning method is applied to accomplish the interactions of customers'decision-making behaviors.Then,the muti-agent simulation model is implemented on the software Anylogic,considering local net-work effects and designing network evolution rules.At last,the co-evolution rules of social network and consumer's decision under different network environments are explored,and the impacts of product utility parameters,communication intensi-ty,the numbers of seed users on the co-evolution and the marginal time of link-broken are analyzed in the process of new product diffusion.The Research shows that:① even though the original structures of network are different,the relations between consumers are reorganized in a new way with higher aggregation and stability,and the co-evolution is accelerated and positive to each other in small-world network and scale-free network;②even though the co-evolution outcomes show dual character under the scale-free network with a high degree,the variation tendencies of network structure are highly similar in the process of the diffusion;③the product value increases with the promotion of network effect,while the posi-tive functions of the co-evolution will eliminate,if the network effect is extremely high or low;④the diffusion outcome will not be influenced when the scale of seed users reach up to a certain Extend;⑤ the marginal time of link-broken will not have an influence on the outcome of the diffusion on the condition of the co-evolution achieving a stable state.