华南理工大学学报(自然科学版) ›› 2019, Vol. 47 ›› Issue (7): 105-111.doi: 10.12141/j.issn.1000-565X.180116

• 交通运输工程 • 上一篇    下一篇

基于 K 均值聚类与随机森林算法的居民低碳出行意向数据挖掘

吴文静1 景鹏2† 贾洪飞1 张铭航1   

  1. 1. 吉林大学 交通学院,吉林 长春 130000; 2. 江苏大学 汽车与交通工程学院,江苏 镇江 212013
  • 收稿日期:2018-03-14 修回日期:2019-03-27 出版日期:2019-07-25 发布日期:2019-06-01
  • 通信作者: 景鹏(1978-),男,博士,副教授,主要从事交通运输系统规划、交通仿真研究. E-mail:jingpeng@ujs.edu.cn
  • 作者简介:吴文静(1980-),女,博士,副教授,主要从事交通运输系统规划、交通行为分析研究. E-mail:wuwj@ jlu. edu. cn
  • 基金资助:
    教育部人文社会科学研究项目(19YJCZH189)

Low Carbon Travel Intention Data Mining for Residents Based on K-means Clustering and Random Forest Algorithm

 WU Wenjing1 JING Peng2 JIA Hongfei1 ZHANG Minghang1   

  1.  1. School of Transportation,Jilin University,Changchun 130000,Jilin,China; 2. School of Automotive and TrafficEngineering,Jiangsu University,Zhenjiang 212013,Jiangsu,China
  • Received:2018-03-14 Revised:2019-03-27 Online:2019-07-25 Published:2019-06-01
  • Contact: 景鹏(1978-),男,博士,副教授,主要从事交通运输系统规划、交通仿真研究. E-mail:jingpeng@ujs.edu.cn
  • About author:吴文静(1980-),女,博士,副教授,主要从事交通运输系统规划、交通行为分析研究. E-mail:wuwj@ jlu. edu. cn
  • Supported by:
    Supported by the Research Project on Humanities and Social Sciences of Ministry of Education(19YJCZH189) 

摘要: 对居民低碳意识的形成机理进行研究,可以为交通管理者引导城市居民选择低 碳出行方式提供重要依据. 运用数据挖掘技术对低碳出行问卷数据进行分析;将计划行为 理论框架下的 15 维问题视为表征居民低碳出行意愿的内在原因变量,应用 K 均值聚类 算法对居民低碳出行意愿强度进行归类,并将所得结果作为被解释变量应用于随机森林 模型中,探讨居民的社会属性特征、出行特征等对其低碳出行意愿的作用机理. 结果表明: 基于 Silhouette 指标检验及 t-SNE 降维,居民低碳出行意愿可划分为 3 类:强烈、中立、不 强烈;基于重要性指标显示影响最为显著的 4 项因素分别是居民的职业、居住地、家庭构 成、通勤时间. 研究结果从多个角度为城市交通低碳化发展及管理提供政策建议.

关键词: 低碳出行意愿, 数据挖掘, K 均值聚类, 随机森林, Silhouette 指标检验

Abstract: The research on the formation mechanism of residents' low carbon awareness will provide important basis for traffic managers to guide urban residents to choose low carbon travel. The questionnaire survey data of low car- bon travel intention were analyzed. The 15-dimensional questions under the framework of the theory of planned be- havior were considered as the intrinsic variables for residents' low carbon travel intention. K-means clustering algo- rithm was used to classify individual types according to individuals' willingness intensity. The results were used as the explanatory variables in the random forest algorithm for identifying individuals’attribute characteristics and travel characteristics of different types of individuals. The results show that,based on the Silhouette index test and the t-SNE dimension reduction visualization,the residents' low carbon travel intention can be divided into three categories: willingness,neutrality and unwilling; and based on the importance of the indicators,the 4 most signi- ficant factors are identified as the individual's occupation,residence,family composition,travel time. The re- search results provide suggestions for the development and management of urban traffic low carbonization from multi- ple perspectives.

Key words: low carbon travel intention, data mining, K-means clustering, random forest, Silhouette index test

中图分类号: