Mining city-wide encounters in real-time

Abstract

Recent advancements in data mining coupled with the ubiquity of mobile devices has led to the possibility of mining for events in real-time. We introduce the problem of mining for an individual's encounters. As people travel, they may have encounters with one another. We are interested in detecting the encounters of traveling individuals at the exact moment in which each of them occur. A simple solution is to use a nearest neighbor search to return potential encounters, this results in slow query response times. To mine for encounters in real-time, we introduce a new algorithm that is efficient in capturing encounters by exploiting the observation that just the neighbors in a defined proximity needs to be maintained. Our evaluation demonstrates that our proposed method mines for encounters for millions of individuals in a city area within milliseconds.