The Research On Financial Distress Forecasting Based On Nonlinear Combination Model

The research on financial distress is one of hot topics in the area of study on corporatefinance, and financial distress is also considered one of serious threats that manyenterprises are faced with. Especially, with the quickening process of global integration,the competition between enterprises becomes more and more fierce. But the enterprisefalling into the financial distress is a gradual process, and usually most enterprises’financial position goes through the process from the normal to deterioration, and falls intodistress and even goes bankrupt finally. So before enterprise’s financial distress occurs,there is threatened. And financial distress is predictable. Therefore, making timely andeffective financial distress prediction by constructing models has important theoretical andpractical significance.Financial distress is very complicated, which is due to enterprise’s internal factors andexternal environment. The accuracy and stability of adopting a single forecasting modelcan’t be guaranteed. Different forecasting models have their own characteristics. Thecombined forecasting model is to give several different models the corresponding weightsso that they can complement each other. Based on manufacturing listed companies, theassay took special treatment as a symbol of financial distress, firstly analyzed thecharacteristics and the causes of financial difficulties, then selected financial indicators andnon-financial indicators used for early-warning model from solvency, profitability,operation ability, corporate governance structure and eliminated the index which is notsignificant by nonparametric test, then extracted the common factor through Factoranalysis method and established Logic regression, BP neural network as well as supportvector machine (SVM) model for early warning. Their prediction accuracy is84.48%,89.66%and87.93%. Finally, the paper inputted each company’s actual resultinto nonlinear combined forecasting model based on fuzzy neural network. In conclusion,these forecasting results of the model are relatively accurate achieving satisfying effect.