In order to improve the recommendation accuracy, it is important to use a variety of models that compensated for each other's shortcomings. In this paper, we propose a hybrid recommender system based on web page clustering and web usage mining. Firstly, we select significant sentences from web pages. Secondly, we extract features from the significant sentences and construct relevant concepts. Finally we use the similarity of web pages to cluster them into different themes. The different themes imply different preferences. The hybrid approach integrates web page clustering into web usage mining and personalization processes. The experimental results show that the combination of the two complementary models can improve the precision rate, coverage rate and matching rate effectively and also help improve the overall solution.