Evolutionary computation for database marketing
- PDF / 206,984 Bytes
- 10 Pages / 652 x 822 pts Page_size
- 36 Downloads / 206 Views
Siddhartha Bhattacharyya is an associate professor of Information and Decision Sciences in the College of Business at the University of Illinois, Chicago. His research covers data mining, data warehousing and various aspect of business intelligence. His work appears in a range of academic journals and he is a regular speaker in different academic and industry conferences and forums.
Abstract This paper examines the use of evolutionary computation (EC) techiques, like genetic algorithms and genetic programmes, for data mining and, more specifically, to address problems in database marketing. Beginning with a brief introduction to the basics of genetic search, it provides an overview of two general application areas in database marketing where EC offers distinct advantages: the development of models optimised to specific targeting depths, and models that simultaneously optimise on multiple objectives.
Siddhartha Bhattacharyya Information and Decision Sciences, College of Business Administration, University of Illinois at Chicago, 601 S. Morgan Street (MC 294), Chicago, IL 60607, USA Tel: ⫹1 312 993 0534; Fax: ⫹ 312 993 0388; e-mail: [email protected]
INTRODUCTION There has been a growing interest in evolutionary computation (EC) techniques addressing problems across diverse domains. Techniques like genetic algorithms and genetic programming offer a search approach based loosely on principles of natural selection and biological evolution. They provide a powerful, general purpose search mechanism that has found application in problems ranging from the design of engines and aerospace structures, scheduling of manufacturing operations and timetabling to portfolio and investment management, modelling of economic phenomena, incorporating adaptive mechanisms in autonomous agents and other models and, in general, for obtaining solutions for hard optimisation problems that are not amenable to solution using traditional approaches. EC approaches have been applied for
䉷 Henry Stewart Publications 1479-182X (2003)
Vol. 10, 4, 343–352
classification and data mining, and have been noted to offer unique advantages for modelling in database marketing.1–3 These advantages stem from the representational flexibility allowed on the form that a model may take, and from the largely open formulation of the search objective (fitness function). As is well known, models may take a variety of forms based on the nature of the problem and available data, the characteristics in a desirable solution and the modelling technique used. Often the model form is dictated by the specific technique used. For example, logistic regression yields a model for the dependent variable that is functionally linear in the set of predictor variables, while CHAID or CART models take the form of decision trees or restricted rule sets. Model representation is crucial since it largely determines the nature of patterns that are discernible from the
Journal of Database Marketing
343
Bhattacharyya
data. Evolutionary search can be usefully applied with a range of repr
Data Loading...
