Discovering periodic frequent travel patterns of individual metro passengers considering different time granularities and station attributes

Abstract

Periodic frequent pattern discovery is a non-trivial task to discover frequent patterns based on user interests using a periodicity measure. Although conventional algorithms for periodic frequent pattern detection have numerous applications, there is still little research on periodic frequent pattern detection of individual passengers in the metro. The travel behavior of individual passengers has complex spatio-temporal characteristics in the metro network, which may pose new challenges in discovering periodic frequent patterns of individual metro passengers and developing mining algorithms based on real-world smart card data. This study addresses these issues by proposing a novel pattern for metro passenger travel pattern called periodic frequent passenger traffic patterns with time granularities and station attributes (PFPTS). This discovered pattern can automatically capture the features of the temporal dimension (morning and evening peak hours, week) and the spatial dimension (entering and leaving stations). The corresponding complete mining algorithm with the PFPTS-tree structure has been developed. To evaluate the performance of PFPTS-tree, several experiments are conducted on one-year real-world smart card data collected by an automatic fare collection system in a certain large metro network. The results show that PFPTS-Tree is efficient and can discover numerous interesting periodic frequent patterns of metro passengers in the real-world dataset.