METHOD AND DEVICE FOR DETECTING DATA ERROR, SOFTWARE AND STORAGE MEDIUM THEREFOR

摘要

PROBLEM TO BE SOLVED: To provide a method for detecting the error of data on a database at high speed with high efficiency and high accuracy on the database containing at least two kinds of data and containing a correspondent relation capable of classifying a certain kind of classification source data with the other kind of classification destination data. SOLUTION: The classification on the database is captured as a class on a neural network, these classifications are divided into small scale double-class problems, a plurality of modules are configured, and it is operated whether each of modules is converged in the learning process of the neural network or not. When the module is not converged, the presence of error in the correspondent relation in the module is decided and such a module is extracted.