Research On Database Schema Matching Based On BP Neural Network

With the rapid development of information technology and the popularity of the Internet, a lot of data appear in front of people, people can easily access a wealth of information, thus facilitating our lives, learning and work. At the same time, people are increasingly inclined to use the integrated information, but because of different environments and different designs, leading to heterogeneous data sources are mostly between, and among the data sources into one comprehensive utilization of such a heterogeneous information, an obstacle, in order to solve this problem, the need for data integration. The key issue for data integration is to find the correct semantic mapping, that is, pattern matching. Between the heterogeneous data sources to solve the critical operational issues - pattern matching, has increasingly attracted attention and research.After deeply studying and analyzing the current matching algorithm,the attribute matching method based on BP neural network is better than other methods,it can effectively carry out attribute matching.However,the current attribute matching methods based on BP neural network still have deficiencies, such as,too large matching space,too much interfering matching,bad matching for the different types of the same attribute,etc.In response to these shortcomings, this thesis propose a new matching algorithm-HPAM matching algorithm,the algorithm has five phases, including the table of matching stage,the attributes matching of similar tables stage,the remaining attributes matching stage,the additional attribute matching stage,the user authentication phase.Table matching method, parallel way of training and matching method,a bi-directional filtering method and the different types of the same attribute matching method are involved in HPAM matching algorithm.HPAM algorithm improves the efficiency of matching, matching precision rate, matching the rate of recall.Finally,HPAM matching algorithm are verified by experiments,and compare analysis of experimental results,experiments show the superiority of HPAM matching algorithm,and discuss the future direction of further improvement.