Optimization methods in diagnostics problems

Some results of the studies obtained by making use of nonsmooth discriminant analysis (NDA) are discussed. This tool represents the optimization approach. Algorithms based on NDA turn out to be quite competitive and efficient. NDA employs, among others, methods of nonsmooth analysis and nondifferentiable optimization. Natural criterion functionals, their approximations (surrogate functionals) as well as the main expert method are desсribed in the paper. By means of the approach presented, a database of the patients of a psychiatric hospital treated due to schizophrenia is analyzed.