B. The Model Testing Results
In this section, we present the testing results of our neural network diagnostic model as shown in Figure 3 to estimate its probability of misclassification error for malignant breast cancer and benign disease outcomes using the resubstitution method.
When the training model completed its training in Figure 1, the final trained weights from the training model were loaded into the neural network diagnostic model. Using the same pattern dataset including all of the available 830 clinical instances of the mammographic mass data, the diagnostic model was tested to estimate its probability of misclassification error in distinguishing malignant breast cancer and benign disease for breast biopsy outcome predictions. During the testing process, the threshold T of the two-lever hard-limit classifier in Figure 3 was set to 0.5. If the output of the diagnostic model was 1, it indicated malignant breast cancer. If the output of the diagnostic model was 0, it indicated benign disease.