Non-Modellable Risk Factor (NMRF) Measurement Using Gaussian Process Regression (GPR)

One innovation defined in the new market risk rules by the Fundamental Review of the Trading Book (FRTB) is the Non-Modellable Risk Factor (NMRF) framework. This new concept introduces a methodology to differentiate between modellable and non-modellable risk factors in the Internal Models Approach (IMA). The modellability assessment is based on two dimensions: the time observability and the price availability. This study is focused on the modelling approach for NMRFs respecting the regulation requirements. First, we present the FRTB requirements and the European Bank Authority (EBA) model suggested for NMRFs. We then introduce the concept and theory of Gaussian Process Regression (GPR). Thereafter, we show the motivation and we also explain the adequate model for this issue respecting the regulator requirements. Finally, we present results by comparing our model with the EBA one and relate extensions to other issues