개인 GPS 데이터 기반의 주요 장소 및 경로 학습

초록

Mobile devices equipped with various sensors can be exploited for providing context-aware services. One popular form of the context is location. In this paper, we present methods for learning routes between a predefined set of significant locations, i.e., home, workplaces, frequently-used bus stops and subway stations, based on personal GPS data. Learned personal routes among such significant locations of a user could be utilized for predicting the user’s destination when he or she is traveling. Our approach was applied to real GPS datasets from six subjects, consisting of more than 15 million coordinates. The accuracy of our methods for significant location learning was about 80% within a boundary of 30 meters. Our method for destination prediction achieved more than 74% of accuracies after a user completing 50% of the trip.

키워드