A cancer research group plans to use a data set to train a classifier, and to help doctors in their clinical practice to detect whether the patient is suffering from diseases, cancer hospital interns were asked to choose the best performance in the trained model model (many types of classifier has been training, such as KNN and SVM, neural network, etc.), interns use 70% of the data as the training set, the other 20% as a validation set, the last 10% as a test set, the intern training 10 kinds of different model, and test set records the precision of interns in the best performance model accuracy up to 90%, interns come to the conclusion that this model is the best, this view is correct?