A0729
Title: Functional data driven financial risk management: An application to the NASDAQ Index
Authors: Fatimah Alshahrani - Princess Nourah bint Abdulrahman University (Saudi Arabia) [presenting]
Zoulikha Kaid - King Khalid University (Saudi Arabia)
Zouaoui Chikr Elmezouar - King Khalid University (Saudi Arabia)
Ali Laksaci - King Khalid University (Saudi Arabia)
Raja M Almarzoqi - Prince Saud Al Faisal Institute for Diplomatic Studies (Saudi Arabia)
Abstract: A new data-driven approach is introduced to manage financial risk in a real-time context. Inspired by the recent development of big data modelling, new dynamic models of financial risk analysis are developed using the statistical analysis of random curve data. More precisely, market risk is assessed through the Functional Expectile Regression (FER), the Functional Conditional Value at Risk (FCVR) and the Functional Conditional Expected Shortfall (FCES). These functional models are estimated using the nonparametric approach. Because the distribution of the volatility of the financial time series is usually not characterized by a specific distribution, the nonparametric approach is more adequate than the parametric for the financial time series. Thus, combining the functional smoothing of the financial data with the nonparametric adjustment of the risk models allows us to increase the accuracy of the existing risk models that are based on the parametric multivariate approach. This approach has been confirmed by an empirical analysis performed on NASDAQ composite index data. It covers 20 years of the daily return of the NASDAQ market index. A comparative study between the three functional models using different back-testing measures demonstrates that the FER model shows more variability, which allows better detection of the risk in crisis as well as in calm periods.
