Skip to Main Content
In this paper we present the test results of an extension developed to the Hierarchical Binary Decision Tree (HBDT) ensemble classifier that was introduced in the WHISPERS2009 conference. The new methodology is the Probability Membership Value based Ensemble (PMVE). It uses the HBDTC designer algorithm to select suitable processing chains for decision fusion, but the actual decision fusion is carried out in an unsupervised manner using a weighted re-mapping of the probability membership values of multiple classifiers. The test data used in this study was acquired during the “red sludge” event in Hungary at the end of 2010. Results of traditional classification approaches are also presented for sake of comparison.