Predicting Asia sovereign debt crises based on support vector machine

摘要

This paper designs and develops an Asia sovereign debt crises predictive model based on support vector machine. A three year time window is defined to observe debt crises probability. Taking 173 samples from eighteen emerging market countries in Asia, the empirical work reveals that the model can predict sovereign debt crises in next three years with accuracy 89.19%. Data analysis also shows repayment history, reserve/MGS ratio and accumulated arrear ratio are top three factors related to debt crisis status.