Can Link Analysis Be Applied to Identify Behavioral Patterns in Train Recorder Data?

Objective: A proof-of-concept analysis was conducted to establish whether link analysis could be applied to data from on-train recorders to detect patterns of behavior that could act as leading indicators of potential safety issues. Background: On-train data recorders capture data about driving behavior on thousands of routine journeys every day and offer a source of untapped data that could be used to offer insights into human behavior. Method: Data from 17 journeys undertaken by six drivers on the same route over a 16-hr period were analyzed using link analysis, and four key metrics were examined: number of links, network density, diameter, and sociometric status. Results: The results established that link analysis can be usefully applied to data captured from on-vehicle recorders. The four metrics revealed key differences in normal driver behavior. These differences have promising construct validity as leading indicators. Conclusion: Link analysis is one method that could be usefully applied to exploit data routinely gathered by on-vehicle data recorders. It facilitates a proactive approach to safety based on leading indicators, offers a clearer understanding of what constitutes normal driving behavior, and identifies trends at the interface of people and systems, which is currently a key area of strategic risk. Application: These research findings have direct applications in the field of transport data monitoring. They offer a means of automatically detecting patterns in driver behavior that could act as leading indicators of problems during operation and that could be used in the proactive monitoring of driver competence, risk management, and even infrastructure design.