Mining travel patterns from mobile ticket applications

Abstract

Customers’ travel patterns are highly interesting for transportation companies due tothe insight it gives over the use for their services. Logs of the customers’ location datais an important source for such companies. However, such data is not collected and isprivately owned by the individual customers. To find the customers’ travel patterns,their location data requires to match a coded map of transportation network.This paper introduces a novel solution that automatically collects location and timedata from the customers without the need for the customers to actively submit thedata. This paper presents an innovative pre-processing and post-processing of locationdata. The pre-process, cleans and anonymizes the parts of the data that are addressableto the individual customers, and optimizes the data for later processes. The proposedsolution, by using classification techniques, is able to match the customers’ location datato coded transportation networks and yield high accuracy of travel matching, even withanonymized incomplete data. Examples are provided on the location data collected inArendal area, Norway. The matching accuracy of the presented solution is relativelyhigh which is indeed promising results.