Financial Crisis Warning Model Based On Genetic Algorithm And Wavelet Neural Network

Since the Shanghai and Shenzhen stock exchangeestablished for thirty years, our capital market has been booming andprovides strong support to the listed company. However, because of thecompetition environment of market economy becoming more and morefiercely, the capital market met lots of challenges, more and more listedcompanies are faced with serious financial risk, finally become into STplate or forced from the stock market. This situation not only affects thenormal development of the capital market, also give all the stakeholder,such as investors, creditors, etc, brought huge economic losses. Financialcrisis warning is a system which can find out the crisis signal by somesensibility index. The study of financial crisis warning can helpcompanies avoid financial crisis effectively and has very importantsignificance of theoretical and realistic for investors, creditors andgovernment regulators.Currently，scholars from all countries are being studied on how toestablish an effective financial crisis early warning model, and alreadyhave yielded substantial results. In the choice of warning methods, thetraditional statistical methods are more extensive applied, which Logisticregression is most commonly used. But the traditional statistical methodsare based on the variable which must meet hypothesis conditions,therefore, these conditions restrict the use of statistical models. Datamining method overcome the defects of the statistical methods whichhave too strict requirements for sample data, and make the input to theoutput can realize arbitrary nonlinear mapping, so they can establish moreaccurate warning modelBased on the study of foreign scholars, combined the geneticalgorithm with wavelet neural networks and used the model(GA-WNN)to study of financial crisis warning in the listed companies ofmanufacturing industry. This model both advantages of wavelet analysisand ANN so that it will get better prediction effect. At the same time,thecomparison was made about GA-WNN model, logistic model,BP neuralnetwork model and SVM model. The results show that, whether the year of t-2or the year of t-3, the prediction accuracy of GA-WNN modelalways be the highest, general prediction accuracy were97%and88.7%.The result illustrated that the GA-WNN model not only applicableto short-term financial early warning, and at the same time, suitable formedium and long-term forecast, so it has great practical value.