文献类型：会议论文

英文题名：Research on Rail Traffic Safety Factor Model Based on Deep Learning

作者：Zhao, Ping[1]; Sun, Lian-ying[2]; Tu, Shuai[1]; Wang, Jin-Feng[2]; Wan, Ying[1]

第一作者：Zhao, Ping

通讯作者：Sun, Lian-ying

机构：[1] Smart City College, Beijing Union University, Beijing, 100101, China; [2] College of Urban Rail Transit and Logistics, Beijing Union University, Beijing, 100101, China

第一机构：北京联合大学智慧城市学院

会议论文集：Artificial Intelligence and Security - 6th International Conference, ICAIS 2020, Proceedings

会议日期：July 17, 2020 - July 20, 2020

会议地点：Hohhot, China

语种：英文

外文关键词：Accidents - Learning systems - Light rail transit - Long short-term memory - Safety factor - Urban transportation

摘要：With the development of urban construction, rail transit has become the main means of transportation for people’s daily life. How to effectively improve the efficiency of urban rail transit operations and reduce the occurrence of accidents is very necessary to build a model of rail transit safety global factors. For the cases of rail transit accidents, the characteristics are recorded in the form of text. The CNN-LSTM method based on BERT embedding is used to classify the accident cases. The accuracy of the test set is 90.65%. Based on the deep learning method, the urban rail transit accident cases in the past 20 years were analyzed, and the safety graph of the dominant factors affecting the safe operation of rail transit was constructed. The results show that equipment and facilities factors, circuit signals, and connecting lines are the main factors affecting the safe operation of rail transit. Passengers, management and operators are indirect factors that cause train failure. The constructed safety graph provides a reference for urban rail transit operation management, making rail transit safety from passive emergency to fine prevention possible. ? 2020, Springer Nature Singapore Pte Ltd.