Abstract: DataBase Complex Networks (DBCN) is a kind of metric for management information systems, which provides a simplified and visualized description of business logic and a self-introduced documentary. To extract DBCNs, two algorithms are proposed: (1) based on the primary and foreign key associations of database tables algorithm; (2) based on the hidden semanteme associations algorithm and its extension. Through the analysis on 9 software databases, the statistical features of DBCN are investigated and found that in-degree distribution is more disperse than that of out-degree, and tables with greater in-degree are the backbone nodes. In order to accurately construct DBCN to facilitate the software engineering, a group of naming criteria is proposed for hidden semanteme associations.