The original dataset is divided into three subsets: [[training]], [[validation]], and [[testing]]. The training set is used to train the model, the validation set is used to fine-tune [[hyperparameter]] and assess performance, and the testing set is reserved for a final evaluation.