基于偏好学习法的大数据流组合优化多智能体研究  

Research on Big data Combinatorial optimization Multi Agent Based on Preference Learning Method

作　　者：张冠珠[1] 苏杉[1] 赵雪峰[1] 常璐璐[2] ZHANG Guanzhu;SU Shan;ZHAO Xuefeng;CHANG Lulu(Binzhou Polytechnic,Binzhou,ShanDong 256600,China;Binzhou University,Binzhou,ShanDong 256600,China)

机构地区：[1]滨州职业学院,山东滨州256600 [2]滨州学院,山东滨州256600

出　　处：《自动化与仪器仪表》2024年第10期47-51,共5页Automation & Instrumentation

基　　金：2021年滨州市社会科学规划课题(2021-SKGH-67)。

摘　　要：数据的分布和特征可能会随时间变化而变化,导致数据的组合优化聚类能力较差,对此,提出基于偏好学习法的大数据流组合优化多智能体研究。首先进行组合优化多智能体控制约束参数和对象模型的构建。然后,结合群体最佳位置寻优和局部适应度控制方法建立大数据流组合优化多智能体进化的动态寻优模型。最后,采用QoS属性的度量分析和偏好学习方法,构建多智能体进化的交叉变异控制模型,并通过优化调整权重、学习因子和随机数等参数,提高多智能体进化算法在分布式大数据组合优化聚类中的效果,以实现组合优化多智能体进化算法优化。实验结果表明,采用本文方法后的收敛性较好,标准方差较低,控制在1以下,时间开销最大为0.9 s,轮廓系数式中保持在0.8以上,说明本方法有效提高了分布式大数据流的动态监测和挖掘能力。The distribution and characteristics of data may change with time,resulting in poor clustering ability of Combinatorial optimization of data.Therefore,a multi-agent research on Combinatorial optimization of Big data flow based on preference learning method is proposed.Firstly,the Combinatorial optimization multi-agent control constraint parameters and object model are constructed.Then,the dynamic optimization model of Big data flow Combinatorial optimization multi-agent evolution is established by combining the optimal location of the population and local fitness control methods.Finally,the measurement analysis and preference learning methods of QoS attributes are used to build the cross mutation control model of multi-agent evolution,and the effect of multi-agent Evolutionary algorithm in distributed Big data Combinatorial optimization clustering is improved by optimizing and adjusting the weight,learning factor,random number and other parameters,so as to realize the optimization of Combinatorial optimization multi-agent Evolutionary algorithm.The experimental results show that the convergence of the method in this paper is good,the standard deviation is low,controlled below 1,the maximum time cost is 0.9s,and the contour coefficient formula remains above 0.8,which shows that the method in this paper effectively improves the dynamic monitoring and mining capabilities of distributed Big data streams.

关 键 词：偏好学习 组合优化 多智能体 交叉变异控制 聚类 动态寻优 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]