Developing SFNN models to predict financial distress of construction companies

Abstract

This research integrates the concepts of self-organizing feature map optimization, fuzzy, and hyper-rectangular composite Neural Networks (called SFNN) to provide a new method for forecasting corporate financial distress. This method not only offers improved rate of prediction accuracy but also offers rules as a reference for examining corporate financial status. For the considered data sample the method satisfies the criteria for 95% confidence level, 3% limit of error, and 50–50 proportion. A total of 1615 effective financial reports from 42 listed construction related companies over the last decade are collected and analyzed. Each financial report contains 25 ratios that bankers commonly use to grade companies which are set as the input attributes. Following comprehensive descriptions of the three algorithms, the SFNN model is constructed. It achieves 85.1% accuracy for predicting corporate financial distress and 49 and 48 valuable rules, for determining the “failed” or “non-failed” of construction companies. Practitioners may even directly use the rules without running the model as a means of quickly and conveniently examining their corporate financial status.