Efficient Global Programming Model for Discriminant Analysis

Conventional statistical analysis includes the capacity to systematically assign individuals to groups. We suggest alternative assignment procedures, utilizing a set of interrelated goal programming formulations. This paper represents an effort to suggest ways by which the discriminant problem might reasonably be addressed via straightforward linear goal programming formulations. Simple and direct, such formulations may ultimately compete with conventional approaches free of the classical assumptions and possessing a stronger intuitive appeal. We further demonstrate via simple illustration the potential of these procedures to play a significant part in addressing the discriminant problem, and indicate fundamental ideas that lay the foundation for other more sophisticated approaches. KeywordsCredit scoring, mathematical programming, multiple criteria and multiple constraint-level programming, Discriminant Analysis.