Boosted Ensemble of the Nadaraya-Watson estimators in regression task on the boundary of the feature space

Abstract

Nowadays, there is considerable interest in using ensembles learning technique to improve the performance of a single learner. Different ensemble learning methods are used in combination with one of the various single learners to avoid some negative effects of this learner implementation. This paper deals with the boundary problem of the Nadaraya-Watson estimator.The boosted ensemble of the Nadaraya-Watson estimators is proposed to solve this problem. Experimental results showed the high efficiency of ensemble approach in regression task on the boundary as well as in the middle of the feature space.