Active Learning from Stream Data

In this paper, we propose a new research problem on active learning from data streams where data volumes grow continuously. The objective is to label a small portion of stream data from 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 a classifier ensemble’s 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 combined to build an active learning framework for data streams. Index Terms Active learning, classifier ensemble, stream data.