文献类型：期刊文献

中文题名：模糊认知图的算法改进与应用综述

英文题名：A review of algorithm improvement and application of fuzzy cognitive map

作者：李慧[1];陈红倩[2];马丽仪[1];梁磊[1];孙旸[1]

第一作者：李慧

机构：[1]北京联合大学管理学院;[2]北京工商大学计算机与信息工程学院

第一机构：北京联合大学管理学院

年份：2016

卷号：52

期号：4

起止页码：746-761

中文期刊名：南京大学学报：自然科学版

收录：CSTPCD;;北大核心:【北大核心2014】；CSCD:【CSCD2015_2016】；

基金：北京市自然基金(4154066;9164028);教育部人文社会科学研究项目(15YJCZH114)

语种：中文

中文关键词：模糊认知图;演化模型;学习方法;软计算

外文关键词：fuzzy cognitive maps; evolutionary model; learning method; soft computing

摘要：模糊认知图(fuzzy cognitive maps,FCM)是一个推理网络,使用循环有向图对知识进行表示与推理,通过模糊逻辑与神经网络的组合完成建模、分析、决策支持、预测等任务.针对模糊认知图的研究现状,从模型构建与学习算法、结构拓展和领域应用创新三个方面,针对近些年来的模糊认知图的算法改进进行了综述.在模型构建与学习算法方面,针对基于Hebbian的学习算法、基于群体的学习算法、基于Hebbian和演化类型混合的学习算法进行了综述,并对算法的主要特征和适用领域进行了描述.在结构模型拓展方面,重点综述了在有专家参与的决策支持领域、高度不确定的复杂系统、动态的领域环境和实时控制系统领域四种领域环境中,对模糊认知图的演化模型改进,并针对模糊认知图中各改进算法的优缺点进行了比较和综述.最后,综述了模糊认知图在各领域中的应用特点和应用案例.
FCM（fuzzy cognitive map）is an inference network which represents knowledge and reasoning using cyclic digraphs.It can implement a number of tasks such as modeling,analysis,decision making and forecast by combining the fuzzy logic with neural networks.It is reviewed for the improvement of algorithms for FCM in recent years in this survey.The reviewed improvement on FCM algorithm includes model construction and learning algorithms,structure development and field application innovation.In the aspect of model construction and learning algorithms,the Hebbian-based learning algorithm,group-based learning algorithm and hybrid learning algorithm（combination of Hebbianbased and evolution）were reviewed.The main characteristics of the algorithm and application fields are described.In the aspect of structural model development,the improved algorithms of evolution models of four application environment are reviewed.The four application environments include decision support system,uncertainty system,dynamicalsystem and real-time control system.The advantages and disadvantages of each improved algorithm are compared and summarized at last.