Online grnn-based ensembles for regression on evolving data streams

Abstract

In this paper, a novel procedure for regression analysis in the case of non-stationary data streams is presented. Despite numerous applications, the regression task is rarely considered in a scientific literature, e.g. compared to classification task. The proposed method applies an ensemble technique to deal with data streams (especially with concept drift). As weak learners, a nonparametric estimator of regression is used. Every single weak model (weak learner) is able to track a specific type of the non-stationarity. The experimental section demonstrates that the proposed algorithm allows for tracking different types nonstationarities and increases accuracy with respect to a single estimator.