The vast amounts of data are stored in different database management systems, but thedifferences of the database environment and designer lead to the heterogeneity between the datasources, which has formed the so-called "information island". Database integration technology caneliminate the heterogeneity, detect the abnormal and improve comprehensive utilization of theinformation. This paper proposes Web Service Technology and Attribute Matching Technology tosolve the problem of heterogeneous database integration.Web Service Data Integration Technology can optimize the traditional integration methods,make the system has more higher scalability and more real-time performance. The core problem ofData Integration is to realize the semantic mapping relations correctly, that is the Attribute Matching.Attribute Matching is the key to solve the problem of heterogeneity between data sources.This paper studies Attribute Matching Technology in the current existing database integration,Learning-Based Solution is much better than the Rule-based Solution (such as BP neural network)on Attribute Matching problem. But the study finds that the current Attribute Matching methodbased on BP network still exists some shortcomings, such as too much interferential match itemsand too large matching spaces. Aiming at the shortcomings of the current method, this paperimproves the Attribute Matching method by three important indicators-Matching precision、Matching recall、Matching efficiency. We mainly propose a new Attribute Matching Algorithmbased on the BP neural network. The algorithm is called Classification Bi-directional FilteringMatching (CBFM).Finally, the raised CBFM algorithm is verified by the experiments done on Matlab7.0.Comparative analysis of the experimental results shows that the efficiency、precision and recall rate can beimproved obviously by using the CBFM algorithm, which proves the superiority of the algorithm. |