基于文化算法的多agent联盟在网络舆情监测中的应用  

Application of multi-agent coalition in network public opinion monitoring based on cultural algorithm

作　　者：柳赛男[1,2] 陈明亮[1] 

机构地区：[1]浙江大学管理学院,杭州310058 [2]杭州电子科技大学数字媒体与艺术设计学院,杭州310018

出　　处：《控制与决策》2014年第9期1724-1728,共5页Control and Decision

基　　金：国家自然科学基金项目(70971116);浙江省自然科学基金项目(Y6090332)

摘　　要：对网络舆情监测监控云平台上集聚的海量技术agent,如果只是逐一agent搜寻使联盟的利益最大化,将耗费大量的时间和费用.为解决该问题,提出了基于"能力群"的动态联盟机制,通过动态增加或删除"能力群"中的agent来增加联盟完成复杂任务的能力;并提出一种基于文化算法的"能力群"agent联盟算法,详细讨论了算法的流程、编码、选择、交叉和变异操作的规则,以及基于"能力群"特征的交叉算法.仿真结果表明,所提出的算法是可行和有效的.It would waste a lot of time and money to search agent one by one for achieving the best coalition＇s behalf in a cloud platform of network public opinion monitoring. In order to solve the above problem, a dynamic coalition mechanism based on the concept of ＂ability group＂ is proposed. The proposed mechanism can add in new agent or delete old agent from ＂ability group＂ according to the performance of agent. In this way, more complex tasks can be completed by selecting agents from these ＂ability group＂ to form agent coalition. A coalition algorithm based on the culture algorithm is proposed. The calculation process and the rules of encoding, selection, crossover and mutation are discussed in detail. The simulation result shows that the proposed approach is feasible and effective.

关 键 词：网络舆情 多AGENT系统 联盟 文化算法 编码规则 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]