Failure discrimination by semi-definite programming using a maximal margin ellipsoidal surface

Yohei Okada, Hiroshi Konno

Abstract

ABSTRACT

We propose a new approach for failure discriminant analysis. The basic idea of the new method is to separate multi dimensional financial data corresponding to ongoing and failed companies by an ellipsoidal surface which is estimated by using a maximal margin hyperplane. We introduce two kinds of norms to construct the surface and compare the precision of prediction with other models including a semi-definite logit model.