Search efficiency and accuracy of resource are important considerations for search algorithm in peer-to-peer （P2P） network.Most search algorithms use flooding among neighbor nodes to search relevant resource.However,this usually causes great amount of redundant messages,which results in high search costs and low search precision.In this paper,we use vector space model （VSM） and relevance ranking algorithms to construct overlay network,and a novel search mechanism search with K-iteration preference （SKIP） based on semantic group for P2P networks is proposed to efficiently solve these problems.The key idea of SKIP is to reorder the semantic neighbors of nodes according to relevant scores and to utilize preference selection during the process of query.We analysis and implement the scheme and reveal that the SKIP provides a low overhead on topology maintenance,which can be effectively used in P2P searching and verify it outperformance in higher precision and lower search cost by comparing with current semantic-based searching mechanism gnutella-like efficient searching system （GES）.
