This paper tends to provide an answer to a difficult problem: Matching large XML schemas. Scalable Matching acquires a long execution time other than decreasing the quality of matches. In this paper, we propose an XML schema decomposition approach as a solution for large schema matching problem. The presented approach identifies the common structures between and within XML schemas, and breaks down these input schemas. Our method uses tree mining techniques to identify these common structures and to select the most relevant sub-parts of large schemas for matching. As proved by our experiments, the proposed approach improves the performance of schema matching and offers a better quality of matches in comparison to other existing matching tools.