Neural Network Algorithm In The Application Of Heterogeneous Database Attributes Match

In this paper a comprehensive analysis of existing heterogeneous database attributes matching the major issues involved. match the attributes of the tasks and solve the matching problem as a major attribute, the characteristics and shortcomings of current methods Department conducted a study Analysis of a neural network theory to solve the problem of matching attributes of feasibility. Based on further analysis of the current literature and BP neural network methods of matching the attribute, that different input counterparts in the same neural network may affect the output neural network is one of the main factors accuracy of the results. and the specific theoretical and experimental evidence of environmental certification. To address the problem, the papers show that the similarity between the attribute set on the basis of compatibility. Matching data and the exchange of training data, the pros and two neural network training method. Heterogeneous networks and proved positive and negative results with the equivalent output after conversion, Training constructed by several different from the general inspection of the two-way idea, a two-way test. The test proved that the two-way misjudge the probability is less than literature. And the development of a prototype system, in both theory and experiment proved that the realization of the two-way test can interfere with effective filtering of data.