For the demand of scientific and technological researchers engaged in innovative activities which need frequently retrieving patents from database, and the intelligent degree of the current patent databases is not enough high for retrieving, this paper proposed an intelligent recommendation algorithm and introduced the development of the corresponding software. Inputted with the retrieval words, the algorithm outputs the results including not only patents which the retrieval system provides but also a group of recommended patents. The algorithm first explores the correlation among the patents, and then calculates the correlation degree among them, finally sorts all the recommended patents according to the correlation degree in order to form an orderly recommended patents set. Experimental results illustrates that there indeed exits correlation between the recommended patents and the retrieval word.