Abstract: Country risk is one kind of special risk which comes from other countries when taking part in international activities, and it is of great importance to understand the law of its dynamic features. In this paper, considering the complexity and mutability of country risk, a novel research framework which concentrates on multi-scale feature extraction and identification is proposed based on the thought of "decomposition and integration". Firstly, country risk data is decomposed to several intrinsic mode functions. Then these intrinsic mode functions are reconstructed to three components of different scales, which represent high frequency scale, low frequency scale and trend respectively. Furthermore, variance contribution rate, correlation coefficient and the Shapley value are introduced to depict features of dynamic fluctuation, correlations between different scales and original data and the global importance of different scales respectively. At last, empirical experiments are given by selecting OPEC countries as cases. It can be found that modes correlation and variance contribution features can effectively classify the sample countries, and the classified results using the two different features have consistency.The global importance of high frequency scale, low frequency scale and the trend is 1:1:3, that is, global importance, which can be obtained by the Shapley value, is coherence and stable between all the countries. Above all, the framework can not only offer more dynamic information of the country risk, but also can be regarded as a new method for the comprehensive identification, monitory and prediction to the country risk.

Key words: country risk, dynamics, multi-scales, Shapley value, ensemble EMD