Visual- and Model-Based Techniques for Validating Corporate Traffic Information Chain
Kin-Fai Chan, Hans Wüst, Henk van Hulst, Jacorien A. A. Wouters, Nanne J. van der Zijpp, David Koh, Paul Janssen · Transportation Research Board 87th Annual MeetingTransportation Research Board · 2008
One of the objectives of the Department of Transportation (DoT) in the Netherlands is to provide information about traffic conditions to road users and traffic managers. The usefulness of this information depends on the quality of the underlying traffic data. Traffic information is generated in a chain of applications, and in each processing step, data can get lost or get corrupted. Missing data are easy to detect, but corrupt data are not. So far, few applications were known that can check large amounts of data for abnormalities, and make an automated distinction between corrupt data and traffic-related outliers (e.g. because of accidents). The Da Vinci project, as described in this paper, aims at developing an application that can detect corrupt data and assist the specialist in analyzing the source of the error, so that the problem can be solved quickly. Da Vinci includes different validation models which are used together to achieve the desired results. The models have access to the data from each processing station of the information chain. Furthermore, an analyzing module was built to help the specialist making case-specific drill-downs in the underlying data in order to find the cause of the corruption. A prototype of Da Vinci has been operational since spring 2007. It successfully identified causes of different errors and help resolving operational problems which was unsolved for a long time. Different examples are presented in this paper showing how the Da Vinci can be used to improve data quality.