Preprocessing of Samples in Modeling of Fetal Macrosomia with Counter Propagation Neural Network

In modeling of fetal macrosomia, some inconsistent data are mixed in samples. Because the procedure of childbearing has been finished, the sample data can not be validated by visiting previous pregnant woman. In order to eliminate the inaccurate samples, a counter propagation neural network is established. Some traditional methods are also used to verify the discarded samples. The new training set shows better classification than original data set.