Research On DataBase Schema Matching Based On Usage

With the development of the Internet technology and people's increasing needs to the information, the database has become increasingly important resources for information storing and sharing, a large amount of data were stored in kinds of databases after collection and sorting. It's very urgent for different enterprises and departments to share and integrate their data, heterogeneous data sources (relational database, object-oriented database, XML, etc.) problem caused by such different environment has become the main obstacle of the data integration. The primary problem to realize the data integration is to define the semantic mapping relation between database models, so-called the database schema matching. As a result, database schema matching has become the key-link of data integration.After further analysis on existing schema matching method, aiming at the problem of Usage-Based schema matching method only consider the use of the information partly, in order to further improve the matching precision rate and matching the rate of recall, this paper improved the Usage-Based method, and proposed a database schema matching method based on usage. In this method attributes vectors were extracted from a database log first, then the data were preprocess, and were matched depending on the similarity between attributes which were calculated by the genetic algorithm at last. Aiming at the problem that most of the existing schema matching methods ignore the usage of attributes, proposed an schema matching based on the various information.In this algorithm, the data indicators was expended together with the attribute use information, data instances and schema information, and BP neural network structure of the algorithm was proposed.Finally, the feasibility of database schema matching method based on usage mapping algorithm and schema matching based on the various information mapping algorithm as well as the improvement of matching precision rate and matching the rate of recall were proved by experiments.