Identification of Corresponding Numerical Attributes in Heterogeneous Databases Based on Instances

Identification of attributes in heterogeneous databases is widely known as the schema matching problem, and many studies have been published. Although existing studies using schema information have been proposed, schema information cannot always be used. In such a case, a schema matching method using instances is used as an alternative method, which is called instance-based schema matching. In this paper, we propose an instance-based schema matching method for the attributes storing numerical data. It uses data distributions and correlations between two attributes. The proposed method is experimentally evaluated by using several databases. As a result, the proposed method could find appropriate attribute correspondences.