Recent years, many researchers are focus on personalized recommendation in the area of knowledge discovery. Due to the user privacy, a recommender system finds it is hard to help individual user to find the items which the user prefers quickly and precisely. So, we combine domain features with AHP model to help us solve the problem. In this paper, we introduce a new recommendation method named AHP-Based domain recommendation method. It intends to utilize item domain features to construct user preference model, and then generates recommendation results by combining user preference model with collaborative filtering (CF) algorithm. After this, we verify our own method by comparing with other three recommendation methods including items-based, ratings-based and F-MADM. As we suspect, the results show our method achieves the best result.