Knowledge Discovery and Data Mining

International Journal of Electrical Engineering Education, Oct 2001 by Nedic, Dusko

M. A. Bramer (ed.), Knowledge Discovery and Data Mining, IEE, 1999, 308 pp., L48; $84.

Our capability for generating and collecting data has increased rapidly in the last decades of the 20th Century. This has resulted in an enormous growth in the amount of stored data. To cope with this growth, sophisticated relational and object-oriented database management systems have been created. However, storing and organising these vast amounts of data is not enough. Techniques to exploit and extract useful knowledge from this mass of data must also be developed.

This book offers an examination of principles and methods in the rapidly advancing field of knowledge discovery and data mining. Written from both a theoretical and a practical perspective, it will be useful for most people who are interested in these topics. Each chapter is designed as a stand-alone guide to an important topic and offers welldescribed algorithms and interesting examples. For example, the chapter by Liu and Cheng discusses the identification of outliers. The statistical treatment of these outliers is discussed in some depth and offers new tools to add to the panoply of descriptive statistics.

Extracting time-related knowledge using data mining techniques is a major challenge for researchers. The chapter by Chen and Petrounias defines this problem and proposes a framework for temporal data mining. Relational databases are the main source of input data for various data mining techniques and several chapters in the book illustrate how data mining techniques can be integrated with relational databases using structured query languages. In addition the chapter by Shapcott, McClean and Scotney demonstrates how to use background knowledge about relational databases. Improved performance can then be achieved through database redesign and the application of logic programming and integrity constraints.

The chapter by Brierley and Batty will be of particular interest to electrical engineers since it provides a well-written introduction to data mining with neural networks, using patterns of electricity consumption as an example. Beginning from the basic principles of neural networks, back-propagation algorithms and load modelling, these authors present a comprehensive survey of the application of neural networks to load forecasting.

Dusko Nedic UMIST