11.1 Introduction

It is estimated that across the world data storage doubles every 18 months (Milne et al., 1997). Sources range from retail (Song et al., 2001), to the telecommunications industry (Hatonen et al., 1996), to industrial processes (Kresta et al., 1991). Although conceivably containing valuable information, this data is largely archived, with its full potential never realised. Data mining is the non-trivial process of identifying valid, novel, potentially useful, and ultimately understandable patterns in data. Applications within diverse areas such as marketing, finance and industrial processes have been considered, with varying degrees of success. However, despite being a widely applied technique, it is said that three-quarters of all companies who attempt data mining projects fail to produce worthwhile results (Matthews, 1997). Unfortunately, this only proves that the potential of data mining, with regard to the available data, is often greater than the reality.

The most successful applications are probably in the fields of scientific research and industrial processes, e.g. chemometrics and chemical engineering (MacGregor et al., 1994; Dunia et al., 1996; Chen and Liao, 2002), industrial process control (Milne et al., 1997; Sebzalli et al., 2000) and to a lesser degree power engineering, where data mining has been used, for example, for fault analysis in transmission networks (Wehenkel, 1996; Rayudu et al., 1997). The emphasis in these applications has been the provision of on-line support systems for operators, in the form of fault detection and diagnosis, and high-level analysis of system operation and performance...