Extracting Information from Sequences of Financial Ratios with Markov for Discrimination: An Application to Bankruptcy Prediction

In this paper, we propose a method that extracts information from sequences of financial ratios and investigate the usefulness of this information for bankruptcy prediction, which constitutes an important class of financial services. We use the annual financial reports available from an external financial information services provider to extract predictors based on the Markov for Discrimination (MFD) methodology. These predictors are used as inputs in a binary classification model, which applies logistic regression to estimate the odds of bankruptcy. The results suggest that MFD-based predictors can achieve substantial predictive performance in terms of the AUC and the 5-percent predictive lift, which are two relevant performance metrics in our case.