An Overview of Classifier Fusion Methods

Abstract

A number of classifier fusion methods have been
\nrecently developed opening an alternative approach
\nleading to a potential improvement in the
\nclassification performance. As there is little theory of
\ninformation fusion itself, currently we are faced with
\ndifferent methods designed for different problems and
\nproducing different results. This paper gives an
\noverview of classifier fusion methods and attempts to
\nidentify new trends that may dominate this area of
\nresearch in future. A taxonomy of fusion methods
\ntrying to bring some order into the existing “pudding
\nof diversities” is also provided.