Traffic delay detection by mining ticket validation transactions

One of the important aspects of traffic management systems is their ability to detect traffic incidents. In this paper, we propose a method for mining the ticket validation timestamps for detecting the delays in public transportation vehicle ridership. We establish a number of criteria crucial for building a robust detector. Method we propose mines the user travelling information for the normal and abnormal patterns in bus ridership. The method observes the transfer time in different places along the routes, and selects such places in an optimal way. Aggregating both normal and abnormal patterns in multi-variate data stream allows to build a reliable detector of traffic incidents. The feasibility and effectiveness of the proposed method is tested on two travelling datasets collected in Nancy and Lyon (France) public transport systems.