Оценивание кредитоспособности предприятий в условиях скрытой марковской зависимости рейтингов

The Markov-swithing multivariate linear regression model for the problem of companies’ credit ratings estimation is introduced. On the assumption of hidden Markov dependency of classes of states the maximum likelihood estimates for the model has been derived. For the model’s parameters and classes of states estimation the expectation-maximization and discriminant analysis algorithms are proposed