Extraction of User Context for Recommendation from Access History
Takaaki Idera, Kenji D. Nakamura, Shigeru Oyanagi · 2015
Recommendation system is widely used to provide favorite information for a user. Current works show that the quality of recommendation can be enhanced by using user context such as time, location and situation. However, preparing enough amount of contextual information is not easy. The method to generate user context systematically is neccesary. This paper proposes a method to extract user context from calendar information and WWW access history for recom- mending restaurants. Five elements of user context such as time, location, situation, category, and budget are extracted systematically. The proposed method uses SVM(support vevtor machine) to associate access history into situation. The accuracy of the proposed method is evaluated from various aspects by experiment. The result shows the effectiveness of this approach. The proposed system is implemented on the Android, and shows the useful recommendations for users.