Research On Financial Distress Early-Warning Of Listed Companies Based On Ternary Financial Positions

Now stock market provides more financing channels for companies, with opportunities and challenges being created. Enterprises with good operation and management seize the opportunity, showing increasingly good momentum of development, while enterprise faced with the plight will be more difficult. As we know, the complex economic environment changes every moment, in order to regulate the operation of the market, the SEC takes some actions, extremly to treat special treatment for enterprises who have been in loss-making state for at least two consecutive years. However there are some enterprises who are enduring short-term losses or being in a poor performance status. Though not been spcial treated, they are already at the edge of the crisis. So using historical financial data to reflect business management activities and to offer warning point to enterprises and other parties have far-reaching significance. Previous studies are mostly concentrated in the area of ST and non-ST samples. Companies in the intermediate gray area with unstable financial state do not attract enough attention. Under the guidance of the prudence principle in accounting, this paper separates these precarious companies alone and classes the whole financial state into three categories, that is, financail crisis(outright failture), financial unstable and financial good. In the following empirical research, firstly, financial indicators for the model are screening from six aspects: profitability, solvency, cash flow ability, development ability, operation ability and shareholder profitability. Then it is the prepation for the model. All the indicators are tested through K-S, remaining those who can make a difference among different types. Heatmap figures display the correlations between indicators, principle component analysis solves this correlated problem, forming new unrelated component factors explaining the complete information. Following the preparations, the extracted new component factors are applied in the latest main methods in the field of financial crisis early warning to forecast the financial state of companies. The methods include artificial neural network method, support vector machines, K nearest neighbor, AdaBoost, random forests. The results show that, ANN, under the natural rate of sample state, the predicted effect of the three classifications are less than ideal. In contrast, the predicted effect of SVM is relatively accurate, it’s rate can reach90%. Adaboost, K nearest neighbor, random forest, their classification results on three of the company’s financial position in the natural state is perfect, namely the use of historical data of these methods can accuratly forecasts the financial position of the year.