Training trees on tails with applications to portfolio choice

摘要: In this article, we investigate the impact of truncating training data when fitting regression trees. We argue that times can be curtailed by reducing sample without any loss in out-of-sample accuracy as long prediction model has been trained on tails dependent variable, is, ‘average’ observations have discarded from sample. Filtering instances an features are selected to yield splits and help reduce overfitting favoring predictors with monotonous impacts variable. test technique exercise portfolio selection which shows its benefits. The implications our results decisive for time-consuming tasks such hyperparameter tuning validation.