Dynamic Weighted Fusion of Adaptive Classifier Ensembles Based on Changing Data Streams

摘要: Adapting classification systems according to new input data streams raises several challenges in changing environments. Although adaptive ensemble-based strategies have been proposed preserve previously-acquired knowledge and reduce corruption, the fusion of multiple classifiers trained represent different concepts can increase uncertainty prediction level, since only a sub-set all classifier may be relevant. In this paper, score-level technique, called $S_{wavg_h}$ , is where each dynamically weighted similarity between an pattern histogram representation concept present ensemble. During operations, Hellinger distance every previously-learned computed, score resemblance underlying distribution. Simulation produced with synthetic problems indicate that technique able system performance when incorporate abrupt changes, yet maintains level comparable average rule changes are more gradual.