Advanced Textile Technology ›› 2024, Vol. 32 ›› Issue (6): 108-115.

Previous Articles     Next Articles

Mining the industry chain link relationship of clothing enterprises based on the industry chain map

  

  1. 1.School of Computer Science and Technology (School of Artificial Intelligence), Zhejiang Sci-Tech University, Hangzhou 310018, China; 2. Zhejiang Modern Textile Technology Innovation Center (Jianhu Laboratory), Shaoxing 312000, China;3. Zhejiang Science and Technology Project Management Service Center, Hangzhou 310018, China
  • Online:2024-06-10 Published:2024-06-17

基于产业链图谱的服装企业产业链挂链关系挖掘

  

  1. 1.浙江理工大学计算机科学与技术学院(人工智能学院),杭州 310018; 2.浙江省现代纺织技术创新中心(鉴湖实验室),浙江绍兴 312000;3.浙江省科技项目管理服务中心,杭州 310018

Abstract: The construction of the clothing industry chain map has become a focal area and a key strategy for the digital upgrade of China's clothing industry. Serving as a vital tool, the clothing industry chain map helps enterprises and researchers better understand and grasp the structure, relationships, and dynamics of the entire industry chain.
This study aims to use digital technology to visually present the entire clothing industry chain, enhancing efficiency, reducing costs, optimizing resource allocation, and ultimately boosting the competitiveness and market position of enterprises.In the clothing industry chain map, determining which industry chain point a company belongs to and understanding the relationships among various enterprises is crucial for industry investment, resource optimization, improved production efficiency, and reduced costs. However, traditional methods of enterprise linkage often involve manual examination of company names, business scopes, and product information, leading to time-consuming and inefficient processes with suboptimal results. Therefore, researching automatic linkage algorithms for clothing enterprises is of practical significance and theoretical value in optimizing the clothing industry chain map.Current research efforts are primarily focused on text information mining and machine learning methods. However, limited research has been conducted on how to use the enterprise profiles and industrial chain map for automatic linkage in the industry chain. This study addresses this gap by collecting enterprise information, extracting keywords, establishing an enterprise information database, and proposing an automatic linkage algorithm based on the CoSENT model. The algorithm utilizes the CoSENT model to calculate the similarity between enterprise keywords and industry chain points, filters matching results through custom rules, assesses the relevance between keywords and points, and achieves automatic linkage in the industry chain for enterprises. Leveraging machine learning technology, this approach provides a more feasible solution for handling vast amounts of information related to clothing enterprises.
Experimental results demonstrate that the proposed algorithm significantly outperforms other traditional algorithms on the F1-Measure metric. Compared to the Jaccard method, the accuracy of this algorithm improved by 14%; compared to the Word2Vec method, it improved by 10.5%; and compared to the SBERT method, it improved by 2.5%. This substantial enhancement elevates the accuracy and efficiency of enterprise linkage, providing robust support and guidance for optimizing the clothing industry chain map. Future research directions include collecting more enterprise information to build richer enterprise profiles, further enhancing linkage efficiency. This study offers a practical solution for the digital upgrade and optimization of the clothing industry chain.

Key words: garment industry chain, industrial chain map, automatic linkage algorithm, CoSENT model

摘要: 服装产业是全球最重要的制造行业之一,而服装产业链图谱则是服装产业生态中的重要工具。为了服装相关企业能快速、准确挂链,文章研究并构建服装产业链图谱,将产业链中的链点、关系和属性进行建模和表示,再通过企业信息收集和企业关键词提取构建企业信息数据库,从而提出了一种产业链企业自动挂链算法。该算法基于CoSENT模型计算企业关键词和产业链链点之间的相似性,并通过自定义规则对匹配结果进行过滤,进而评估关键词和链点之间的相关性,自动匹配和选择最优的产业链图谱链点,实现企业的产业链自动挂链。通过与其他匹配算法的对比实验表明,该算法在F1-Measure指标上明显优于其他算法(比基于Jaccard方法高14%,比Word2Vec方法高10.5%,比SBERT方法高2.5%),显著提升了企业挂链效率和准确性,为优化服装产业链图谱提供了有力的支撑和参考。

关键词: 服装产业链, 产业链图谱, 自动挂链算法, CoSENT模型

CLC Number: