Output details

A weighted voting framework for classifiers ensembles

<24>The importance of this work is in its theoretical nature. It is well known that the overwhelming amount of the classifier ensemble research is driven by empirical studies, while theory is rare and very welcomed. Here we propose a theoretical framework which joins together three of the most popular classifier ensemble combination methods (combiners): the majority vote, the weighted majority vote and the Naive Bayes. As a result of developing the framework, we discovered a new combiner (which we named "Recall") whose logical place is between the weighted majority and the Naive Bayes.