Schema integration plays a central role in numerous database applications, such as Deep Web, DataSpaces and Ontology Merging. Although there have been many researches on schema integration, they all neglect user preference which is a very important factor for improving the quality of mediated schemas. In this paper, we propose the automatic multi-schema integration based on user preference. A new concept named reference schema is introduced to represent user preference. This concept can guide the process of integration to generate mediated schemas according to user preference. Different from previous solutions, our approach employs
-measure and “attribute density” to measure the similarity between schemas. Based on this similarity, we design a top-
ranking algorithm that retrieves
mediate schemas which users really expect. The key component of the algorithm is a pruning strategy which makes use of Divide and Conquer to narrow down the search space of the candidate schemas. Finally, the experimental study demonstrates the effectiveness and good performance of our approach.