A Framework of User Model Based on Semi-Supervised Techniques

With the exponential increase of the information resources on the Web, the need to mine useful information in the personalization system has become more and more important. There are several problems in current personalization applications, which can be solved well with our state-of-the-art user model framework. Active learning strategy is used to obtain more accurate labeled examples as well as semi-supervised machine learning techniques are used to mitigate user human labor. A new profile space which takes contextual information into account can take full advantage of the information in the userpsilas transactional histories. The realization of automatic personalization is more simple and efficient.