Verification of Bayesian Clustering in Travel Behaviour Research _ First Step to Macroanalysis of Travel Behaviour

4th International Conference on Civil and Environmental Engineering for Sustainability

地球科学;生态环境科学;经济学

Satra, P.^1 ; Carsky, J.^1

Department of Transportation Systems, Faculty of Transportation Sciences, Czech Technical University in Prague, Konviktska 20, Praha 1

CZ - 110 00, Czech Republic^1

关键词: Bayesian clustering;    Bayesian statistics;    Clustering process;    Data preprocessing;    Graphic representation;    Model parameters;    Setting parameters;    Travel behaviour;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/140/1/012067/pdf
DOI  :  10.1088/1755-1315/140/1/012067

学科分类：环境科学（综合）

来源: IOP

【 摘 要 】

Our research is looking at the travel behaviour from a macroscopic view, taking one municipality as a basic unit. The travel behaviour of one municipality as a whole is becoming one piece of a data in the research of travel behaviour of a larger area, perhaps a country. A data pre-processing is used to cluster the municipalities in groups, which show similarities in their travel behaviour. Such groups can be then researched for reasons of their prevailing pattern of travel behaviour without any distortion caused by municipalities with a different pattern. This paper deals with actual settings of the clustering process, which is based on Bayesian statistics, particularly the mixture model. An optimization of the settings parameters based on correlation of pointer model parameters and relative number of data in clusters is helpful, however not fully reliable method. Thus, method for graphic representation of clusters needs to be developed in order to check their quality. A training of the setting parameters in 2D has proven to be a beneficial method, because it allows visual control of the produced clusters. The clustering better be applied on separate groups of municipalities, where competition of only identical transport modes can be found.