基于知识图谱的地铁行车应急预案推荐系统的研究与构建

Research and Construction of Subway Operation Emergency Plan Recommendation System Based on Knowledge Graph

摘要: 传统地铁行车调度应急预案依赖人工操作和经验判断，缺乏自动化和智能化的支持现代技术如物联网、大数据分析等可能未被充分利用，导致应急响应速度慢。针对传统应急预案存在的问题，提出地铁行车调度应急预案推荐系统的研究。研究由两个核心模块构成：应急预案知识图谱的构建和应急预案推荐系统的构建。应急预案知识图谱构建任务中，Qwen2.5大语言模型进行应急预案文本知识抽取，Neo4j知识储存和可视化展示，用户可根据需求使用Cypher语句查询所需应急预案知识。意图识别任务中，对文本使用BERT-TextCNN模型进行分类操作，使得自然语言与预定义模板相互匹配，Cypher在应急预案知识图谱进行搜寻相应板块内容，最终输出用户所需信息。

Abstract: Traditional subway operation dispatch emergency plans rely on manual operations and empirical judgment, lacking automated and intelligent support. Modern technologies such as the Internet of Things (IoT) and big data analysis may not be fully utilized, leading to slow emergency response speeds. Aiming at the problems of traditional emergency plans, this paper proposes a study on the subway operation dispatch emergency plan recommendation system, which consists of two core modules: the construction of an emergency plan knowledge graph and the development of an emergency plan recommendation system. In the task of constructing the emergency plan knowledge graph, the Qwen2.5 large-language model is used for knowledge extraction from emergency plan texts, Neo4j for knowledge storage and visual display, and users can use Cypher statements to query the required emergency plan knowledge according to their needs. In the intention recognition task, the BERT-TextCNN model is used for text classification to match natural language with predefined templates. Cypher is applied to search for corresponding sections in the emergency plan knowledge graph, and finally, the information required by users is output.