PDF pdf 1.0MB

Reinforcement Learning In Multi-Factor Real Options Modelling

Previously, we introduced a methodology based on exercise boundary fitting (EBF) in an effort to develop a practical Monte Carlo simulation-based real options approach. Our theoretical and numerical presentation of the EBF method shows how real world complexity can be overcome through the use of Monte Carlo simulation and that the EBF methodology is very tractable in an industry setting for it is simple enough for managers to understand, yet can account for important real world factors that make the real options model suitable for valuation. However, we recognize that in an effort to account for real world complexities, multiple stochastic factors will need to be modelled. In such cases, the exercise boundaries will be multi-dimensional hyper surfaces. Modelling such surfaces will have its own challenges and will further tax the optimization required with the EBF method. A promising solution to the problem may be the use of reinforcement learning (RL) which we explore in this paper.