Data Mining and Knowledge Discovery 1, 225–231 (1997) c  1997 Kluwer Academic Publishers. Manufactured in The Netherlands. Brief Application Description Visual Data Mining: Recognizing Telephone Calling Fraud KENNETH C. COX kcc@research.bell-labs.com STEPHEN G. EICK eick@research.bell-labs.com GRAHAM J. WILLS gwills@research.bell-labs.com Bell Laboratories/Lucent Technologies, Room 1G-351, 1000 East Warrenville Road, Naperville, IL 60566 RONALD J. BRACHMAN AT&T Laboratories Editor: Received April 2, 1997; Revised April 23, 1997; Accepted Abstract. Human pattern recognition skills are remarkable and in many situations far exceed the ability of automated mining algorithms. By building domain-specific interfaces that present information visually, we can combine human detection with machines’ far greater computational capacity. We illustrate our ideas by describing a suite of visual interfaces we built for telephone fraud detection. Keywords: visualization, fraud, information discovery, interaction, telecommunications data mining 1. Introduction One way to improve our ability to make use of large, complex, information-rich data sets is to build on the fact that people are at the heart of the data mining enterprise. Human pattern recognition skills are remarkable, and far exceed the ability of any existing technology to detect interesting patterns and relevant anomalies. In particular, by properly taking advan- tage of peoples’ abilities to deal with visual presentations, we may revolutionize the way we understand large amounts of data. Researchers at Bell and AT&T Laboratories have worked to exploit the pattern detection capabilities of the human visual system by building a suite of tools and applications that flexibly encode data using color, position, size, and other visual characteristics (Eick and Fyock, 1996). Multiple different views of the same data can be interlinked, so a change in one view is reflected instantaneously in the others. Applications using this technology have been used to solve a wide variety of business-critical problems, including: • improving programmer productivity by visualizing change patterns in large software systems (Ball and Eick, 1996);