Efficiently explaining the predictions of a probabilistic radial basis function classification network

Article type: Research Article

Authors: Robnik-Šikonja, Marko | Štrumbelj, Erik^{; *} | Kononenko, Igor

Affiliations: Faculty of computer and information science, University of Ljubljana, Tržaska, Ljubljana, Slovenia

Correspondence: [*] Corresponding author: Erik Štrumbelj, Faculty of computer and information science, University of Ljubljana, Tržaska 25, 1000 Ljubljana, Slovenia. Tel.: +386 1 4768459; Fax: +386 1 4768498; E-mail: erik.strumbelj@fri.uni-lj.si.

Abstract: A probabilistic radial basis function (PRBF) network is an effective non-linear classifier. However, similar to most other neural network models it is non-transparent, which makes its predictions difficult to interpret. In this paper we show how a one-variable-at-a-time and an all-subsets explanation method can be modified for an equivalent and more efficient use with PRBF network classifiers. We use several artificial and real-life data sets to demonstrate the usefulness of the visualizations and explanations of the PRBF network classifier.

Keywords: Data mining, model visualization, model interpretation, feature importance, neural nets

DOI: 10.3233/IDA-130607

Journal: Intelligent Data Analysis, vol. 17, no. 5, pp. 791-802, 2013