Overfitting and underfitting
Within the context of SML, we will prepare our models by fitting them with historical data. The process of fitting a model generally outputs a measure of how well the model generalizes to data that is similar to the data on which the model was trained. Using this output, usually in the form of precision, accuracy, and recall, we can determine whether the method we implemented or the parameters we changed had a positive impact on our model. If we revisit the definition of ML models that from earlier in this chapter, we specifically refer to them as models that learn or generalize from historical data. Models that are able to learn from historical data are referred to as well-fitted models, in the sense that they are able to perform accurately on new and unseen data.
There are instances in which models are underfitted. Underfitted models generally perform poorly on datasets, which means they have not learned to generalize well. These cases are generally...