A Proposed Risk Modeling Shift from the Approach of Stochastic Differential Equation towards Machine Learning Clustering: Illustration with the Concepts of Anticipative & Responsible VaR
28 Pages Posted: 21 Sep 2017
Date Written: September 18, 2017
The aim of this technical document is threefold with the bigger picture being to contribute, within the challenging regulatory environment, to bring closer together traditional conflicting practices such as trading vs risk as well as risk responsiveness vs stability. In order to achieve this goal, we first expose some of the complexity associated to the risk factors and arbitrage constraints associated with the options and the high frequency markets by re-introducing the Implied Volatility Parametrization (IVP), and the High Frequency Trading Ecosystem (HFTE). The exposed complexity is then contrasted with the current obsolete Risk Methodologies which are based on simplistic SDEs which we extent using the Cointelation model in order to partially address some of the complexity introduced by the challenging regulatory environment such as scenario coherence. We then present a simple Machine Learning clustering methodology which is designed to address and mirror the enhancements of these SDEs in a simpler fashion. We illustrate our findings by introducing few new risk concepts such as the Anticipatible VaR which aims at being a leading as opposed to a lagging (Responsive) risk measure to a market regime change.
Keywords: Stochastic Differential Equation, Gaussian Process, Cointelation, Value at Risk (VaR), Responsive VaR, Stable VaR, Responsible VaR, Anticipative VaR, Anticipatible VaR, Stochastic Differential Equations (SDE), Implied Volatility Parametrization (IVP), High Frequency Trading Ecosystem (HFTE), VAR
Suggested Citation: Suggested Citation