主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2017, Vol. 25 ›› Issue (12): 59-67.doi: 10.16381/j.cnki.issn1003-207x.2017.12.007

• 论文 • 上一篇    下一篇

集群行为中的复杂网络结构合并优化方法

张锴琦1,3, 杜海峰2,3, 王晶晶2,3   

  1. 1. 长安大学经济与管理学院, 陕西 西安 710064;
    2. 西安交通大学公共政策与管理学院, 陕西 西安 710049;
    3. 西安交通大学公共管理与复杂性科学研究中心, 陕西 西安 710049
  • 收稿日期:2016-04-14 修回日期:2016-10-17 出版日期:2017-12-20 发布日期:2018-02-10
  • 通讯作者: 杜海峰(1972-),男(汉族),四川巴中人,西安交通大学公共政策与管理学院,执行院长,教授,博士生导师,研究方向:社会网络分析、系统复杂性及系统优化,E-mail:haifengdu@mail.xjtu.edu.cn E-mail:haifengdu@mail.xjtu.edu.cn
  • 基金资助:

    国家自然科学基金资助项目(71071128);国家社会科学基金重点项目(12AZD110);中央高校基本科研业务费专项资金项目(西安交通大学基本科研业务费青年教师跟踪支持计划);国家社会基金资助项目(17BJY139)

An Optimization Method for the Complex Network Structure Combination in Collective Behavior

ZHANG Kai-qi1,3, DU Hai-feng2,3, WANG Jing-jing2,3   

  1. 1. School of Economics and Management of Chang'an University, Xi'an 710064, China;
    2. School of Public Policy and Administration of Xi'an JiaoTong University, Xi'an 710049, China;
    3. Center for Administration and Complexity Science of Xi'an JiaoTong University, Xi'an 710049, China
  • Received:2016-04-14 Revised:2016-10-17 Online:2017-12-20 Published:2018-02-10

摘要: 集群行为中的人际互动关系可以通过复杂网络结构予以表示,以特定结构特征为目标的结构合并优化问题对于研究集群行为具有很好的应用价值。本文通过数理模型分析,以最小平均路径长度为优化目标,以不同连边策略为变量构建了复杂网络合并优化模型,提出了启发式遗传算法的网络结构合并优化SCOA算法。最后,针对不同复杂网络模型进行实验与仿真,讨论集群行为群体合并的一般策略与相关特征。研究发现,不同的连边规则对网络结构合并性能有显著影响,网络结构合并优化问题客观存在;本文提出的SCOA算法优于现有的同配性连边规则、异配性连边规则与随机连边规则;其所得的连边规则是一个包含同配连边规则与异配连边规则的混合连边规则。本文研究为集群行为中解决群体隔离问题提供了新的分析思路与方法。

关键词: 集群行为, 复杂网络, 网络结构, 结构优化, 结构合并算法

Abstract: The interpersonal relationship in collective behavior can be denoted as the complex network structure. Changes of network structure will influence characteristics and functions of network. Such changes, with the characteristics and functions of network structure as optimal objectives, can be defined as the optimization problem of network structure. The combination of network structure is a common way of network changes, and the combination with a certain kind of structure characteristics is a practically significant optimization problem. Through the description of the mathematical model on the optimization problem of combination of network structure, the optimization problem with the average path length is solvable and sociologically significant. Thus' based on the genetic algorithm, a corresponding optimization is proposed on the combination of network structure. It is found through the experiment of the combination of random network models that the different connected rules have a marked impact on the performance of the combination of network structure, and the optimization problem of the combination of network structure does exist. While the results of proposed algorithms in this paper prove to be better than that of the existing connection rules, additionally, the connected rules of proposed algorithm is a mixed connection rule consisting of both assortative connection rule and disassortative connection rule.

Key words: complex network, network structure, collective behavior, structure optimization, structure combination algorithm

中图分类号: