Learning User Perception to Traveler Situation Awareness Alerts on Mobile Devices

The use of mobile devices to deliver traveler information such as situation awareness and navigation guidance is on the rise. However, an optimal experience – and even more widespread use – may be impeded by the lack of personalized, user-tailored applications. The issue is that users react differently to traffic information, and information perceived useful by one user may be considered as nuisance by another. The authors provide evidence of this from a pilot field test performed on a situation awareness application that alerts the user of approaching slow traffic 1.6 km ahead. The authors furthermore propose a machine learning algorithm based on a support vector machine that segregates alerts into “favorable” vs. “nuisance” based on user feedback and user GPS trace analysis. Their preliminary findings reveal that for 50% of users participating in the experiment, 80% or more of the alerts perceived as nuisance could have been suppressed.