基于BP神经网络与GIS的出行方式识别  

Travel mode identification based on BP neural network and GIS

作　　者：王杰[1] 杨坤[1] 张文斌 WANG Jie;YANG Kun;ZHANG Wenbin(School of Transportation and Vehicle Engineering,Shandong University of Technology,Zibo 255049,China)

机构地区：[1]山东理工大学交通与车辆工程学院,山东淄博255049

出　　处：《山东理工大学学报(自然科学版)》2025年第4期9-15,共7页Journal of Shandong University of Technology:Natural Science Edition

基　　金：山东省重点研发计划项目(2018GGX105010)。

摘　　要：现代城市交通的规划建设离不开大量、详实调查数据的支撑,而现有的问卷调查和出行日志调查内容繁琐、数据更新周期长、效率低下、成本高昂。本文利用GPS采集出行轨迹数据,基于BP神经网络模型结合GIS地图匹配的方法,实现了对步行、自行车、电动自行车、公交车与出租车5种出行方式的识别。采用混淆矩阵验证算法识别精度,结果表明对5种出行方式的识别准确率均达到了95%以上。该方法可为交通出行调查方法的改进提供参考。The planning and construction of modern urban transportation heavily rely on extensive and detailed survey data.Existing questionnaire surveys and travel log surveys are cumbersome,have long data update cycles,low efficiency,and high costs.This study uses GPS to collect travel trajectory data and combines a BP neural network model with GIS map matching to identify five travel modes:walking,cycling,electric vehicles,buses,and taxis.The accuracy of the proposed algorithm is validated through the confusion matrix,and the results show that the recognition accuracy for all five travel modes exceed 95%.The findings provide valuable insights for improving transportation survey methods.

关 键 词：出行方式 识别 BP神经网络 GIS GPS 

分 类 号：U491.1[交通运输工程—交通运输规划与管理]