Computational marketing using an AppleSeed cluster

Computational marketing can be defined as the application of computational science techniques to modelling and understanding market behaviour. Computational science techniques are able to simulate phenomena that are too complex to be explored by existing forms of theory. The techniques are also able to extend the benefits of experimental science into phenomena that are not suited to traditional experimentation, such as the study of social and market processes. This paper reports two activities that provide examples of research using a computational marketing approach. The first example describes work we are currently undertaking with a data mining technique, rough clustering, which is based on rough sets theory. The second example outlines a more ambitious research program that we are also actively pursuing. This is the use of agent based social simulation (ABSS) techniques in modelling consumer markets. Agents can be defined as autonomous software entities that possess an individual representation of the environment with which they interact, a set of goals they wish to achieve in that environment, and various decision making capabilities. They are also capable of communicating with other agents existing in the environment. Both examples of computational science techniques involve numerically intensive computing which requires considerable processing power. Previously, such processing power was only possible using specialised computers (such as the Cray T3E or IBM SP2) housed in a centralised supercomputer centre. Such platforms are expensive, require support staff to maintain them, and are often hard to access for many researchers. The paper describes a comparatively inexpensive alternative using an AppleSeed cluster, based on a network of Macintosh G4 computers connected by an Ethernet switch. The cluster provides a relatively simple to operate “desktop supercomputer” allowing more ambitious data analysis and market modeling projects to be attempted. The traditional theoretical and experimental approaches to science have been joined by a third, the computational approach. Computational science is able to extend the benefits of experimental science into new phenomena that are not suited to traditional experimentation, such as the study of social and market processes (Kohler & Gumerman, 2000; Krischke, 1999). It also offers new forms of theory formulation (Gropp, Lusk & Skjellum, 1994). The growth of this third approach has been stimulated by the rapid development of computing hardware and methodologies since the computational approach’s beginnings in the 1940s. The development of this computational approach can currently be seen in a wide range of academic disciplines. As computational science becomes adopted into each new discipline, a new sub-discipline develops. There is now a emerging literature in computational biology and environmental studies (Deadman, 1999; Rouchier, Bousquet, Barreteau, Le Page & Bonnefoy, 2001; Sumpter & Broomhead, 1998), computational geography (Auerswald & Kim, 1995; Besussi Cecchini & Rinaldi, 1998; Wu & Webster, 1998), computational organizational theory (Butel & Watkins, 2000; Chang & Harrington, 2000; Lomi & Larsen, 1996; Prietula, Carley & Gasser, 1998), computational