Efficient event detection by exploiting crowds.

Encouraging users to participate in community-based sensing and collection for the purpose of identifying events of interest for the community has found important applications in the recent years in a wide variety of domains including entertainment, transportation and environmental monitoring. One important challenge in these settings is how significant events can be detected by exploiting the data sensed, gathered and shared by the crowd, while respecting the resource costs. In this paper we investigate the use of dynamic clustering and sampling techniques that allow us to significantly reduce utilization costs by clustering low-level streams of events based on their geo-spatial locations and then selectively retrieving the ones that depict the highest interest. Our experimental results illustrate that our approach is practical, efficient and depicts good performance.