Solving a large scale semi-definite logit model

Abstract

This paper is concerned with an algorithm for solving a large scale semi-definite logit model which cannot be solved by an outer approximation (cutting plane) algorithm proposed earlier by one of the authors. Outer approximation algorithm can solve a problem with up to 10 financial attributes and 7,800 companies which is less than satisfactory from the viewpoint of failure discriminant analysis. The new algorithm can generate an approximately optimal solution for problems with over 14 attributes and 8,000 companies, by which the quality of failure discriminant analysis would be substantially improved.