Web User Profiling Based on Non-violative Model

This paper proposes a new non-violation based user profiling method. The non-violative strategy is presented to protect the personal privacy and “double identifier" method is used to realize the reuse of profiles. The Markov user model is constructed to understand the user's behavior automatically. Some parameters in the model, such as transition probability matrix and mean holding time matrix are defined to describe the characters of web user dynamically. The approach is implemented in a sample site to help the design and performance evaluation of the website. The test result is satisfactory.