摘要
以往对产业集群的相关实证研究存在数据获取困难、数据维度片面、传统复杂网络理论分析方法可拓展性差等问题.针对以上问题,本文以互联网上的大量非结构化数据为基础,采用图嵌入模型提取集群网络特征的向量空间分析方法,利用互联网公开数据构建产业集群关联网络,结合企业行业分类标准与分析目的设计部分节点标签,使用关系型图卷积神经网络模型(R-GCNs),从产品关联层面进行产业集群特征学习.根据产业集群内企业的嵌入表示和地理位置信息,提出了集群网络嵌入应用分析方法.通过对宁波地区制造业集群相关数据进行实验分析和论证,验证了图嵌入分析方法在量化分析产业集群关联网络特征上的有效性.
Previous research on industrial clusters has problems such as the difficulty in data acquisition or one-sidedness of data dimensions,and poor scalability of analysis method in traditional complex network theory.To address these problems,a vector space analysis method is proposed,which is based on a large amount of unstructured data on the Internet,and employs a graph embedding model to extract the features of cluster associated network.The public data from Internet are used to construct industrial cluster associated network,and part of node labels are designed based on standards of industry classification and purposes of analysis,with which the model of relational graph convolutional networks(R-GCNs)is used to learn the characteristics of industrial clusters from the level of product association.According to the embedding and geographic location information of enterprises in the industrial cluster,the application method of cluster network embedding is also proposed.Through the experimental analysis and demonstration of the relevant data of the manufacturing industry cluster in Ningbo,the effectiveness of the graph embedding analysis method in the quantitative analysis of the characteristics of the industrial cluster associated network is verified.
作者
牛士银
余军合
战洪飞
王瑞
刘东洋
NIU Shiyin;YU Junhe;ZHAN Hongfei;WANG Rui;LIU Dongyang(Faculty of Mechanical Engineering&Mechanics,Ningbo University,Ningbo 315211,China)
出处
《宁波大学学报（理工版）》
CAS
2023年第5期29-36,共8页
Journal of Ningbo University：Natural Science and Engineering Edition
基金
国家重点研发计划项目(2019YFB1707101,2019YFB1707103)
国家自然科学基金(71671097)
浙江省省属高校基本科研项目(SJLZ2023001)
浙江省公益技术应用研究计划项目(LGG20E050010,LGG18E050002)。
关键词
产业集群
关联网络
关系型图卷积网络
区域特性
industrial cluster
associated network
relational graph convolutional networks
regional characteristics
