Detalhes do Documento

With the growth and development of data, the empirical evidence supporting a link between the distance metrics that are used in the instance-based algorithms and generalization has been mounting. In this paper, we look at distinct similarity measures to study its impact on the performance accuracy of incremental instance-based algorithms in pattern recognition problems. An in-depth analysis of the results of the proposed study for a variety of classi cation tasks (binary and multi-way) from various di erent domains shines light on the trade o between the distance metrics and yielded accuracy.