Forecasting Value-at-Risk of Crude Oil Using a Hybrid Arima-Svr-Pot Model

Measuring the value at risk(VaR) of crude oil futures poses a challenge for investors due to the high volatility of these prices. Describing the return in the tail distribution is crucial because these extreme values can trigger larger price fluctuations and market risks. Based on the autoregressive integrated moving average (ARIMA), support vector regression (SVR), and peak over threshold (POT) method from extreme value theory, we have constructed a hybrid model ARIMA-SVR-POT and compared it with the ARIMA-EGARCH model and the ARIMA-SVR model. We selected WTI crude oil futures as the sample, covering the period from June 23, 2016, to September 30, 2022, which includes many extreme events, and conducted empirical analysis. We employed the Kupiec test to assess the effectiveness of these three models in measuring the risk value of crude oil futures returns and reached the following conclusions: the hybrid model ARIMA-SVR-POT predicts the returns and volatility of crude oil prices more accurately than the ARIMA-SVR and ARIMA-EGARCH models. Furthermore, the ARIMA-SVR-POT performs well in capturing the extreme tail of returns. Back-testing results demonstrate that ARIMA-SVR-POT can pass the Kupiec-test at confidence levels of 95%, 99%, 99.5%, and 99.9%, with the LR statistic close to 0. Empirical results further illustrate that our proposed model reflected potential losses in estimating VaR, with the actual loss occurrence probability being closer to the predicted loss probability, indicating superior performance compared to traditional statistical models. The ARIMA-SVR-POT hybrid model presented in this study accurately captures the risk value of crude oil futures price series, providing valuable insights for risk management for crude oil futures investors and companies