主权债务危机早期预警系统研究——基于BP神经网络方法

文章运用BP神经网络方法，构建了主权债务危机早期预警系统。通过对1991-2006年54个发展中国家宏观经济及债务状况数据的实证研究发现，该系统能对未来三年内出现的主权债务危机事件起到较好的预警作用，预警总体效果达86.7%。同时，与二元Logistic模型进行预测对比，发现运用人工神经网络方法对主权债务危机进行预警比二元Logistic方法具有相对优势。

A Research on Early Warning System of Sovereign Debt Crisis: Based on BP Neural Networks

This paper has developed an early warning system of sovereign debt crisis based on BP neural networks method. By using data on macro-economy and debt of 54 developing countries from 1991 to 2006, this paper reveals that the early warning system can effectively predict sovereign debt crisis in the following three years and its overall effect is 86.7%. Furthermore, ANN has an advantage over binary Logistic model in the prediction of sovereign debt crisis.