中文摘要：

应用支持向量机（support vector machine,SVM）多分类方法解决财务的多类预警问题,将以往简单的两类预警模式（正常和危机）扩展到3类分类预警。为提高模型效率及预测准确率,利用主成分分析法对财务指标降维后作为输入样本,运用粒子群算法（particle swarm optimization,PSO）对SVM模型中的参数进行优化,并将建立的PSO-SVM多分类预警模型对200家上市公司进行实证研究。结果表明,该方法有效、可行,较传统SVM及判别分析模型具有更好的预测能力,为企业财务的动态预警提供了新的途径。

英文摘要：

Support vector machine （SVM） classification method was applied to solve multiple financial problems of many types of early warning, and early warning to the previous simple model of two types （ normal and crisis） was extended to three categories of early warning. To improve the efficiency and prediction accuracy of the model ,the principal component analysis of the financial indicators was used as the input samples after dimensionality reduction;the particle swarm optimization （PSO） of the SVM model was used to optimize the parameter; and the PSO - SVM multi - classification of early warning model was established for the 200 listed companies in empirical research. The results show that this method is more effective, feasible, and has better predictive ability for early warning than the traditional SVM model and discriminant analysis. It provides a new dynamic way for the company financial warning.