Abstract
The modeling of user's behavior pattern for personalized information services in mobile environment has recently become a popular research theme. Most of the researches aim at predicting user's future behavior (and/or location) by extracting frequent patterns from the history of location data sequences. However, sometimes user's behavior changes according to the external information such as date, time, weather etc. and we can not accurately predict it based on the location data sequences only. In this paper, we propose a new prediction method including date and time as external information. First the user's travel history (location, date, time) is stored. Then, from the external information, time/date categories that have correlation to the user's destination based on entropy are selected. Using the time/date categories, a destination which depends on the external information is successfully predicted. An application of the method to a data collected from a car navigation system showed an improved performance comparing to the conventional prediction methods. Higher destination prediction accuracy during the first several minutes after user's departure was reported.
