Tail Risk Assessment Using Support Vector Machine

Abstract

In this paper, authors apply Support Vector Regression (SVR) tool Oracle DM in forecasting volatility of Belex 15 index and estimation of Value-at-Risk (VaR). VaR is calculated using SVR model and compared to the results achieved implementing Markov Regime Switching model VaR and Feed Forward Neural Network VaR (ANN FFNN VaR). The results show that the SVR tools give better estimates of VaR comparing to other methods.

Keywords

• Support Vector Regression
• Value at Risk
• Markov Regime Switching