Learning-based user modeling approach in intelligent user interface

A new learning-based user modeling approach is proposed.An implicit,distributed user model can be built by this method.Machine learning is used in two stages.First,learning-based CNP(Contract Network Protocol) is used to get the initial user model.Second,observation-based learning is used to update user model.Compared with other user modeling,this method does not limit the location of user login while it can make full use of user's history information;It dose not violate user's privacy;It possesses dynamic adaptation with respect to preferences of different users,and time-varying preferences of an individual user during a session.