Research On The Listed Company's Finance Crisis Prediction Based On SVM

A listed company is the basis of society economy, its financial status and performance affect government's macro-economic policies and capital market directly. The process that a listed company caught in financial crisis is a step-by-step one, just worsening from regular financial status gradually, and reaching financial crisis ultimately. This indicates that the enterprise financial crisis has its own portents, which are divinable. Therefore, building an accurate and effective listed company financial crisis early warning model has important significance, because it is beneficial both to protecting the investor and the creditor's interests, and helping the operator avoid financial crisis and government monitor the quality of listed companies and the risk of securities market.This paper sums up the characteristics, causes and portents of a listed company's financial crisis on the basis of other scholars'studying both in domestic and abroad, which is helpful to building up a more pragmatic, rational and scientifically listed companies'financial crisis forecasting model. A financial crisis prediction model works by means of an early warning system. Therefore, we need to adopt a reasonable early warning system in order to identify financial crisis before it happens and warn the administration. Support vector machine is a new technique of data mining, which is regarded as the best theory aimed at solving the problem of classification and regression of small sample pool at present. The introduction of support vector machine into listed company financial crisis early warning definitely contributes to the study of listed company financial crisis early warning analysis method. This paper introduces the principles of support vector machine, and has built up a financial crisis early warning model based on support vector machine. This model is then used in progress studies making use of the listed company financial data announced in an open manner. SVM has avoided the artificial neural network's problems such as difficult to ascertain structure, over-study and not study enough. Research result indicates that SVM-based listed company financial crisis early warning model, compared with other forecasting models, has the superiority in forecasting accuracy under small sample data pool condition.