The popular WWW pages are stored in the places, which are close to the users, by WWW Cache technology to speed up fetching these objects. The information in the WWW Cache shows the users? recent interest. The users? interest can be widely used in customizing the WWW pages, filtering the information, pre-fetching the information and so on. The key to how to use the information in the WWW Cache effectively is to build an adaptive user interest model and to construct an adaptive algorithm for interest mining. The interest can be specialized by the tuple (term, weight) in the simple interest model. The association relations are not mined in this model, and then the interest can not be associated in expressing the users? interest. Basing on analyzing the WWW Cache model, we bring forward a two-dimension interest model. It is proved by the real time that the mined users? interest can show the users? current interest states. The inferential relations are well considered in the two-dimension interest model. This model is not the simple extension of the simple interest model, but the full upswing of the model and the related algorithm. In this paper, the interrelated methods on how to store the two-dimension interest model, how to compute the model effectively and how to update the model in real time.