Influence-Aware Predictive Density Queries Under Road-Network Constraints

Density query is a very useful query type that informs users about highly concentrated/dense regions, such as a traffic jam, so as to reschedule their travel plans to save time. However, existing products and research work on density queries still have several limitations which, if can be resolved, will bring more significant benefits to our society. For example, we identify an important problem that has never been studied before. That is none of the existing works on traffic prediction consider the influence of the predicted dense regions on the subsequent traffic flow. Specifically, if road A is estimated to be congested at timestamp t(1), the prediction of the condition on other roads after t(1) should consider the traffic blocked by road A. In this paper, we formally model such influence between multiple density queries and propose an efficient query algorithm. We conducted extensive experiments and the results demonstrate both the effectiveness and efficiency of our approach.