After building a model, it is necessary to find metrics to measure the goodness of it. The model’s performance is generally monitored on new instances that were not a part of the training data.