摘要
为了从大规模GPS轨迹数据中提取出行行为建模所需的必要信息,文章将贝叶斯网络应用到GPS数据处理过程中,建立了出行方式识别的贝叶斯网络模型以个体出行者作为研究对象,以智能手机采集的轨迹信息作为数据源,利用K2算法学习贝叶斯网络结构,采用极大似然估计法学习贝叶斯网络参数。以建立的贝叶斯网络模型为基础,推断了样本的出行方式,实现了步行、自行车、电动车、公交车和小汽车共五种出行方式的自动化识别研究表明,贝叶斯网络适用于出行方式识别研究,且低速点比例和平均方向改变两个指标可以有效提高出行方式识别准确度.
To extract necessary information for investigating travel behavior from a large quantity of track data generated in household travel surveys,a Bayesian network was developed and applied to process the data.Taking individual travelers as objects and track data recorded by smartphones as data sources,we identified travel modes by learning Bayesian network structure using K2 algorithms and estimating network parameters with maximum likelihood methods.We additionally automatically derived travel modes for trips by walk,bicycle,e-bicycle,bus and car by means of the Bayesian network erected in this paper.Results from the study demonstrate that Bayesian network is suitable for travel mode detection and that rate of points with low speed and average direction change provide an opportunity to increase detection accuracy of travel modes.
出处
《统计与决策》
CSSCI
北大核心
2017年第6期75-79,共5页
Statistics & Decision
基金
国家自然科学基金资助项目(51478266)
关键词
贝叶斯网络
结构学习
参数学习
出行方式
Bayesian belief network
structure learning
parameter learning
travel mode
