Assessing Risk in Discrete Event Simulation by Generalized Deviation

Abstract

This paper introduces a generalized deviation concept, inspired by quantitative finance. Standard risk metrics like volatility or downside risk are deconstructed into five general sub-functions for reference, selection, penalization, normalization and re-dimensioning. The advantage of this approach is its flexibility, allowing modeling a wide range of risk perceptions in numerous application fields of discrete event simulation. Several further risk types like upside risk, outside risk, transition risk, critical state risk or countermovement risk are describable and embeddable as special cases of generalized deviation. These risk types are presented with respect to motivation, specification of relevant generalized deviation components, description of application classes, application examples and graphical illustrations. In particular, various options for determining reference states, reference selection and penalty functions are discussed. Implementation features of the generalized deviation metric in the discrete event simulation framework DESMO-J are outlined. Moreover, possible structural extensions as well as additionally implementable risk types are delineated, indicating further application potential and the flexible scope of this approach. It is proposed to complement descriptive standard statistics in discrete event simulation domains by additionally employing risk measurement in terms of generalized deviation as explicated here, to facilitate assessment of undesired simulation dynamics in various application fields.