Forecasting With Ratio Based Multivariate Discriminant Analysis

Abstract

Financial ratio based multivariate discriminant analysis (MDA) has been used as a tool for business forecasting for long time. In this article, we study some of the key issues of MDA in context of bankruptcy prediction, including reducing number of Financial ratios used in the model, the selection of optimal accounting review intervals and noise contamination. Experiments are designed for testing ratio sensitivity and the stability of the MDA models and illustrated using both traditional and eCommerce bankruptcy data. To provide some empirical insights for applications of these models, observations from these experiments are summarized and carefully documented.