Risk-Reward Analysis in Stochastic Dynamic Programming

Stochastic dynamic programming models are extensively used for sequential decision making when outcomes are uncertain. These models have been widely applied in different business contexts such as inventory control, capacity expansion, cash management, etc. The objective in these models is to deduce optimal policies based on expected reward criteria. However, in many cases, managers are concerned about the risks or the variability associated with a set of policies and not just the expected reward. Considering risk and reward simultaneously in a stochastic dynamic setting is a cumbersome task and often difficult to implement for practical purposes. Here we develop heuristics that systematically track the variance and the average reward for a set of policies, which are then utilized to construct efficient frontiers. We apply our heuristics to the inventory control model. Our heuristics perform creditably in providing efficient risk-reward curves. (