中文摘要：

基于新兴市场国家主权债务危机数据,使用Logit模型和支持向量机方法,旨在构建具有较强预测能力的主权债务危机预警系统。实证结果表明,利息与外债占比、距离上次危机年份、实际GDP增长率、到期本息与储备比以及公外债负债率等五个变量,对下一年主权债务危机的发生有很好的预警作用。同时与其他同类研究对比发现,该系统在预警准确率、一类错误率和二类错误率上具有比较优势。而且,使用支持向量机方法在提高预测准确率方面也有较明显的优势。

英文摘要：

Based on the data of sovereign debt crises in the emerging market countries, this study employs the Legit model and the Support Vector Machine method to construct a sovereign debt crises warning system with higher forecasting capacity. The empirical result shows that the five variables, the proportion of interest and external debt, the years from the last crisis, the real GDP growth, the ratio between the principal due plus interests and the reserve, and public external debt ratio, can fair- ly play a warning role in forecasting the next occurrence of sovereign debt crisis. Compared with similar studies, this system has clear advantages in the prediction accuracy, the error rate of the first category and the error rate of the second category. Besides, the use of the Support Vector Machine method also has clear advantage in the improvement of prediction accuracy.