Int. J. Computational Intelligence Studies, Vol. 1, No. 2, 2009 163
Copyright © 2009 Inderscience Enterprises Ltd.
An experimental analysis of the impact of accuracy
degradation in SVM classification
D. Malchiodi
Department of Computer Science,
University of Milan,
Via Comelico 39/41, 20135 Milano, Italy
Fax: +39 0250316333 E-mail: malchiodi@dsi.unimi.it
Abstract: The aim of this paper is to analyse the phenomenon of accuracy
degradation in the samples given as input to SVM classification algorithms. In
particular, the effect of accuracy degradation on the performance of the learnt
classifiers is investigated and compared, if possible, with theoretical results.
The study shows how a family of SVM classification algorithms enhanced in
order to deal with quality measures on the available data handles accuracy
degradation better than the classical SVM approaches to classification.
Keywords: data quality; accuracy; data quality in classification; support vector
classification; machine learning; accuracy degradation; computational
intelligence.
Reference to this paper should be made as follows: Malchiodi, D. (2009) ‘An
experimental analysis of the impact of accuracy degradation in SVM
classification’, Int. J. Computational Intelligence Studies, Vol. 1, No. 2,
pp.163–190.
Biographical notes: Dario Malchiodi received a degree in Computer Science
in 1996 and a PhD in Applied Mathematics in 2000, both from the Milan
University, Italy. He is an Assistant Professor at the same university, where he
teaches laboratory of computer programming, advanced programming and
simulation. His main research area range from probability theory and
mathematical statistics to various aspect of computational and machine
learning, with particular emphasis to the effect of data quality on machine
learning problems.
1 Introduction
The ISO 0402-1986 report defines quality as ‘the totality of feature and characteristics of
a product or service that bear on its ability to satisfy stated or implied methods’ (ISO,
1994; Eurostat, 2003). Note how this definition poses equal emphasis on the terms
product and service, where the difference between them refers to the physical/immaterial
dimension, although the field of product manufacturing has faced quality issues since
long time (see e.g., Burr, 1953), whereas similar instances gained importance only
recently w.r.t. service management (Parasuraman et al., 1985). The definition also
implicitly highlights the relation between quality and value of a product or service.
Indeed, economists have focused since more than 30 years on quality issues (Garvin,
1983; Liepins, 1989; Porter and Millar, 1985; Te’eni, 1993) and on the additional costs