The purpose of the model hθ is to predict output values hθ(x) from unseen input values x (the features), using knowledge θ extracted from a set of data for which the output values y are already known.