文章摘要
张鹤鹏,黄达,杜辰,李晓璐,朱广宇.基于移动数据的用户出行方式识别研究[J].交通运输研究,2018,4(6):47-54.
基于移动数据的用户出行方式识别研究
User Travel Mode Recognition Based on Mobile Location Data
  
DOI:
中文关键词: 移动终端  位置数据  出行方式  城市规划  决策树  C4.5算法
英文关键词: mobile terminal  location data  travel mode  urban planning  decision tree  C4.5 algorithm
基金项目:国家自然科学基金项目(61872037);国家自然科学基金项目(61833002);深圳市交通公用设施建设项目(BYTD-KT-002-2)
作者单位
张鹤鹏 北京交通大学交通运输学院 
黄达 北京交通大学交通运输学院 
杜辰 北京交通大学交通运输学院 
李晓璐 1. 北京交通大学交通运输学院2. 北京交通大学综合交通运输大数据应用技术交通运输行业重点实验室 
朱广宇 1. 北京交通大学交通运输学院2. 北京交通大学综合交通运输大数据应用技术交通运输行业重点实验室 
摘要点击次数: 101
全文下载次数: 118
中文摘要:
      为研究智能手机所采集到的位置数据在识别用户出行方式领域的应用,首先,比选出速度、速度的百分位数、轨迹点数量占比、出行距离、停止率这5 个适用于移动终端定位数据区分出行方式的特征变量,并对各特征变量的判别阈值进行了定义。然后,针对基站分布导致的数据偏差和位置信息漂移等问题,采用扇形定位结合地图匹配技术对数据进行了修正,进而在对时间阈值和距离阈值分割的基础上提出了移动终端用户出行链的获取方法。接着,建立C4.5 决策树模型,以此判别移动终端用户的出行方式。最后,基于在某地区采集的7 000 部移动终端的位置数据(包含:终端编号、定位时刻、经度、纬度) 来对用户的出行方式进行研究。结果表明,模型在区分机动车和非机动车时准确率较高,达到了90%以上;在进一步区分中,如区分步行与自行车以及公交车和小汽车的出行上准确率相对较低,但也达到了80%以上的精度。
英文摘要:
      In order to study the application of location data of intelligent mobile phones in the field of identifying the users’travel modes, firstly, five characteristic variables applied to distinguish travel modes by mobile terminal positioning data were compared and selected, including speed, percentile of speed, proportion of track point number, trip distance and stopping rate. The threshold value of each characteristic variable was determined. Then, aiming at the problems of data deviation and position information drift and other problems caused by base station distribution, the data was modified by sector positioning and map matching technology. Based on the segmentation of time threshold and distance threshold, the acquisition method of mobile terminal user travel chain was proposed. After that, C4.5 decision tree model was established to distinguish mobile terminal users’travel modes. Finally, the location data, including terminal number, positioning time, longitude and latitude, of 7 000 mobile terminals in one area was collected to study those users’travel modes. The results showed that there was high accuracy in distinguishing motor vehicles and non-motor vehicles, which was more than 90%. Then in further distinguish, such as between walking and cycling, as well as bus and car travel, there was relatively low accuracy, but it also achieved an accuracy of more than 80%.
查看全文   查看/发表评论  下载PDF阅读器