Learning Significant Locations from GPS Data with Time Window

摘要

The most important and common service of LBS is to provide the user of his location. In this paper we propose a method to cluster a period of GPS data into meaningful locations by using DBSCAN algorithm. With a time window constraint this method can even distinguish the locations of the same place where the user went at different time. The learned significant locations are important basis of further service, such as predicting the user's future movement, building the user's own map etc. At last, we introduce a prototype system based on the learning method to provide the place information where the user went before and the information is expressed in time-ordered and semantic landmarks directly.