Modern computing systems of all kinds accumulate various data at an almost unimaginable rate. Alongside the advances in technology that make such storage possible has grown a realisation that buried within this mass of data there may exist some knowledge of considerable value. This could be information critical for a company's business success or something leading to a scientific or medical discovery or breakthrough. Most data is simply stored and never examined, but machine-learning technology has the potential to extract knowledge of value (i.e. data mining).
This book considers knowledge discovery - which has been defined as 'the extraction of implicit, previously unknown and potentially useful information from data' - and data mining. Six chapters examine technical issues of considerable practical importance to the future development of this field; issues such as how to overcome feature interaction problems, analysis of outliers, rule discovery, the use of background knowledge, temporal patterns and online analysis processing. There then follow six chapters which describe applications in fields as diverse as medical and health information, meteorology, organic chemistry and the electricity supply industry.
The book grew from a colloquium held in 1998 by the IEE, co-sponsored by the British Computer Society Specialist Group on Expert Systems (BCS-SGES), the Society for Artificial Intelligence and Simulation of Behaviour (AISB) and the International Society for Artificial Intelligence and Education (AIED). The chapters have been expanded considerably from papers presented, and all have been fully refereed.
- Chapter 1: Estimating concept difficulty with cross entropy
- Chapter 2: Analysing outliers by searching for plausible hypotheses
- Chapter 3: Attribute-value distribution as a technique for increasing the efficiency of data mining
- Chapter 4: Using background knowledge with attribute-oriented data mining
- Chapter 5: A development framework for temporal data mining
- Chapter 6: An integrated architecture for OLAP and data mining
- Part II: Knowledge discovery and data mining in practice
- Chapter 7: Empirical studies of the knowledge discovery approach to health-information analysis
- Chapter 8: Direct knowledge discovery and interpretation from a multilayer perceptron network which performs low-back-pain classification
- Chapter 9: Discovering knowledge from low-quality meteorological databases
- Chapter 10: A meteorological knowledge-discovery environment
- Chapter 11: Mining the organic compound jungle - a functional programming approach
- Chapter 12: Data mining with neural networks - an applied example in understanding electricity consumption
University of Portsmouth, UK.
Max Bramer is Professor of Information Technology at the University of Portsmouth, UK. He has been actively involved in research in artificial intelligence since the early 1970s and leads the University of Portsmouth's Artificial Intelligence Research Group. His current research interests include knowledge discovery and data mining and model-based approaches to diagnostic reasoning. He has recently developed the Inducer system, which combines a number of paradigms for the automatic generation of classification rules from examples.
He is Chairman of BCS-SGES and is a member of the lEE's Professional Group on Artificial Intelligence.