< Day Day Up > 

Current computer-based medical diagnostic methods use neural networks, discriminant analysis and other machine learning approaches for medical diagnosis (Doi et al., 1993; Kovalerchuck, Triantaphyllou, Ruiz & Clayton, 1997; Pendharkar, Rodger, Yaverbaum, Herman & Benner, 1999; Wu, Doi, Giger, Metz & Zhang, 1994). Although somewhat useful these approaches do not incorporate the economic considerations of misclassification. There are two types of errors that are encountered in classification systems: false positive (Type I) and false negative (Type II) error. The costs of these errors are not equal. For example, predicting that a patient does not have heart disease when the patient has it is more costly than predicting that a patient has heart disease when he does not have it.

Traditional classification systems such as neural networks and linear discriminant analysis do not allow a user to incorporate a symmetric costs of misclassification. In fact, these costs are considered equal in most machine learning classification systems. In this chapter, we propose and implement a GA-based classification model that allows the decision maker to incorporate misclassification costs. Using simulated, real-life heart disease and liver disorder data sets, we show that the proposed GA model performs better than parametric linear discriminant analysis and a nonparametric linear GA-based model that does not allow decision-makers to incorporate costs.

The rest of the chapter is organized as follows. In "Pure Frontier Models," we provide an overview of linear discriminant analysis and genetic algorithm based models for classification. In "Contrasts of Meaning and Purpose," we suggest modifications to the genetic algorithm based model that incorporates Type I and Type II cost based priorities. "Data Mining Uses and Suggested Guidelines" provides tests of the proposed genetic algorithm model using simulated and real life data sets. The summary of our research and directions for future work are in "Other Models and Applications."

 < Day Day Up > 

Managing Data Mining Technologies in Organizations(c) Techniques and Applications
Managing Data Mining Technologies in Organizations: Techniques and Applications
ISBN: 1591400570
EAN: 2147483647
Year: 2003
Pages: 174 © 2008-2017.
If you may any questions please contact us: