Personalizing threshold values on behavior detection with collaborative filtering

Abstract

We are developing a system which assists users by collaboration between the users and environment. Our collaboration system provides services according to user behavior proactively in homes when environment detects high-level user behavior such as "leaving the home". To realize such a collaboration system, this paper proposes a method for detecting high-level user behavior. The proposed method dynamically sets values suitable for individual behavioral pattern of each user to thresholds used for detection. A conventional method determines threshold values common to all users. However, the common values are not always suitable for all users. Our method determines threshold values suitable for a user by utilizing data of other users whose characteristics are similar to the user, with collaborative filtering. © 2008 Springer-Verlag Berlin Heidelberg.