基于居民出行特征的职住地精细化识别
Fine-grained identification method of home-work location based on travel characteristics of residents
黄兴如 1李奕萱 1刘中亮 1冯瀚斌 1王希昭 1闫龙 1胡博文 1李炫孜 1李大中1
作者信息
- 1. 联通数字科技有限公司 数据智能事业部,北京 100010
- 折叠
摘要
为了解决传统职住模型测算规则的单一性和局限性,降低各区域居民用户因作息规律差异或临时性变化而造成的职住地识别误差,创新性提出一种基于不同区域居民出行特征的职住地精细化识别方法.首先,采用"3 min切片"和"角度+驻留时间+连接次数"等多种方式对手机信令数据进行降噪提炼;然后,基于时空约束密度聚类进行驻留点识别分析;最后,根据各城市居民日常出行特征,通过引入加权驻留时长动态更新各城市区域居民用户职住地测算规则,进而精细化识别不同城市用户职住地分布.实验结果表明,所提方法涉及的过程均合理有效,且最终的职住地识别效果要明显优于传统单一职住模型测算规则,适用于同时批量处理多个区域职住地问题,尤其对因突发状况而产生作息时间变化的城市效果更为显著.
Abstract
To address the simplicity and limitations of the traditional home-work model calculation rules and reduce the identifica-tion errors caused by differences in the daily routines of residents in various regions or temporary changes,this study proposed a fine-grained identification method of home-work location based on the travel characteristics of residents in different regions.First-ly,various methods such as"3-minute slicing"and"angle+stay time+connection frequency"are used to denoise and refine the mobile phone signaling data.Then,based on spatiotemporal constrained density clustering,stay points are identified and ana-lyzed.Finally,according to the daily travel characteristics of residents in various cities,weighted stay duration is introduced to dynamically update the home-work calculation rules for residents in different city areas,thereby refining the identification of home-work distribution for users in different cities.Experimental results show that the processes involved in this method are reasonable and effective,and the final home-work identification results are significantly better than those of traditional single home-work mod-el calculation rules.This method is suitable for batch processing of home-work problems in multiple regions simultaneously,par-ticularly for cities where changes in routines are caused by unexpected events.
关键词
信令数据/出行特征/密度聚类/加权驻留时长/职住地识别Key words
cellular signaling data/travel characteristics/DBSCAN/weighted stay duration/home-work location identification引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据