Dynamic Risk Measurement of Financial Time Series with Heavy-tailed: A New Hybrid Approach

Xinxia Yang, Ratthachat Chatpatanasiri, Pairote Sattayatham

Abstract

This paper proposes a new hybrid approach to measure dynamicrisk of financial time series with heavy-tailed distribution. The proposed method,hereafter referred to as NIG-MSA, exploits the normal inverse Gaussian (NIG)distribution to fit the heavy-tailed distribution, and combines the empirical modedecomposition with support vector regression to structure a multi-scale analysis(MSA) methodology. The validity of NIG-MSA method for volatility predictionis confirmed through Monte Carlo simulation. This method is illustrated with anapplication to the risk measurement of returns on  S&P500 index and our resultsshow that the proposed NIG-MSA approach provides more precise value at riskcalculation than the traditional single-scale model.