Once we have built a multi-class classification network, we'll use it to make predictions on new data that wasn't used during training. We'll start by calling the predict
method, which returns the probability that each input data row belongs to each one of the possible classes. Then, we'll use the predict_classes
method to only output the class prediction as an integer, which might be easier to use in a production system.