A Model Validation Method Based on Convolutional Neural Network

Conventional model validation methods analyze outputs similarity between simulation and real world with same inputs. However, it is hard to guarantee the condition in practice. In order to solve the problem, a method based on convolutional neural network (CNN) is proposed, including data preprocessing, activation function, loss function, and optimization algorithm. Meanwhile, a CNN is established for model validation training and test. Finally, a case study of model validation is presented. The result shows that, the method can obtain 98.5% validation accuracy under the condition of same inputs, and can discriminate credibility levels with different inputs as well.