Refinement of data streams using Minimum Variance principle

In this paper, we propose a refined scheme on active learning from data streams where data volumes grow continuously. The objective is to label a small portion of stream data for which a model is derived to predict future instances as accurately as possible. We propose a classifier-ensemble based active learning framework which selectively labels instances from data streams to build an ensemble classifier. Classifier ensemble's variance directly corresponds to its error rates and the efforts of reducing the variance is equivalent to improving its prediction accuracy. We introduce a Minimum-Variance principle to guide instance labeling process for data streams. The MV principle and the optimal weighting module are proposed to be combined to build an active learning framework for data streams. Results and implementation demonstrate that the percentage of accuracy of the Minimum variance margin method is good as compared to other methods.