Cluster-based user segmentation to optimize needs-oriented transport services

Kurzfassung

The trend toward individualization calls for personalized, needs-oriented, and accessible public transport services. Existing mobility services often fail to meet these requirements, as personal and context-sensitive travel needs are considered insufficiently. Therefore, a user segmentation method was developed to analyse the effect of traveller characteristics (e.g., age, gender, public transport subscription) on route selection, referring to a data set about route decisions (N = 433). A three-stage modular process was used to define the segmentation method. In the first module, user travel data was pre-processed and transformed for cluster analysis. This step addressed the challenges of high dimensionality and mixed data types by applying methods such as principal component analysis or factor analysis for mixed data. In the second module, unsupervised machine learning techniques were applied to cluster users based on the extracted components. In the third module, correlation between user groups and route choice were examined. Applying the segmentation method revealed that male and female, but not non-binary, residents of large capitals with a public transport subscription tend to favour faster routes with fewer transfers. Overall, the results make an important contribution to the specification of a segmentation method for personalized routing options in public transport.