Valid prediction intervals for regression problems

Over the last few decades, various methods have been proposed for estimating prediction intervals in regression settings, including Bayesian methods, ensemble direct interval estimation and conformal methods. An important issue is calibration of these methods: generated should a predefined coverage level, without being overly conservative. In this work, we review above four classes from conceptual experimental point view. Results on benchmark data sets domains highlight large fluctuations performance one set to another. These observations can be attributed violation certain assumptions that are inherent some We illustrate how used as general procedure deliver poor results step.