A Study of Financial Distress Prediction based on Discernibility Matrix and ANN

Financial distress prediction is a significant issue that concerns stakeholders of enterprises. Due to the limitation of the samples available, most studies of financial prediction generally catagorize the financial situations of companies by whether they are bankrupt or specially treated in the stock market. In addition, the variables for prediction are determined mainly by subjective judgments. In order to overcome the influence of these limitations, this paper establishes a reduced variable system with sufficient information by rough set discernibility matrix method, and forms a progressive classification of financial situations by hierarchical clustering. These classifications are transformed as the target value of artificial neural network output layer to enhance the interpretation ability of the whole network. Combined with the inputs determined by Rough Set reduction, a neural network model is established to predict the financial distress, especially the turning point of financial degeneration.