A1594
Title: Predicting emerging market sovereign credit spreads with machine learning/data science techniques
Authors: Gary Anderson - CEMAR LLC (United States) [presenting]
Alena Audzeyeva - Keele University (United Kingdom)
Abstract: Standard financial time series regression analysis has previously been applied to construct a collection of linear models for forecasting credit spreads of sovereign debt issued by Brazil, Mexico, Philippines and Turkey for the period just before and just after the Lehman crisis. We apply a variety of Support Vector Machine Regression (SVMR) kernels to develop a family of parsimonious non-linear regression models. We estimate the model parameters using data science motivated cross-validation techniques. We find that these models can significantly outperform the linear models in Diebold-Mariano and Clark-West tests.
