Abstract: The quality and efficiency of existing methods for scientific paper recommendation still need improvement, furthermore, the large-scale user data in virtual network community has provided a new way for carrying out collaborative recommendation. In this paper a new method for scientific paper recommendation is presented, which is based on academic community service system and can fully employ community user information to improve the quality and efficiency of recommendation. The solutions of key problems in this method, including academic community system design, community discover and collaborative recommendation algorithm based on community are given detailed description. The practical application example showed that this method can provide more accurate and credible scientific paper recommendation service for research users.