A framework for CBR development and experimentation with application to medical diagnosis

Abstract

Most of the data mining techniques focus on classification tasks and so several tools have been designed to help engineers and physician in building new classification systems. However, most of the tools does not take into account the application domain. In this paper we present a case-based reasoning (CBR) tool with the aim to support the development and experimentation of a particular medical classification task: diagnosis. Our tool, eXiT*CBR provides navigation facilities to the engineers and physician across several experimentation results, assures experiment reproducibility and incorporates functionalities for independizing data, if required. The use of eXiT*CBR is illustrated with a Breast Cancer diagnosis system we are currently developing. © 2009 Springer-Verlag Berlin Heidelberg.