Figure 3.

Comparison of the MSE curve trend of control model (left) and new model (right). In both graphs, the horizontal axis indicates the number of iterations during training; the vertical axis shows the model error (MSE) values. Blue curve represents the MSE of the training model in the neural network building process, green curve indicates the MSE trend under data cross-validation; red curve represents the test data (sampled from the neural network training set by removing the extraordinarily good or poor models as outliers).