Stochastic Dynamic Programming (SDP) is an optimization methodology that considers the variability of the system. SDP is studied broadly and applied to various systems, from manufacturing to transportation. For them to work correctly, the SDP models need to parametrize accordingly. Each application counts with different parametrization methodologies. Also, when the number of states increases, the algorithms are inefficient, and the optimal policy is hard to interpret for real-case scenarios. Therefore, the structural properties of the models are studied, so optimal strategies are proposed. In this talk, I will cover SDP applications in Revenue Management for the airline industry and healthcare applications.
For revenue management, we will talk about how to define a promotion policy, which has become standard practice in the airline industry as a strategy to boost the total revenue. Diversion of demand from the regular fare to the markdown price, also called dilution, is a side-effect of offering promotions, which needs to be considered in designing successful campaigns. The performance of this model is examined in two cases from a Latin American airline and demonstrates considerable savings by applying the proposed optimal policy versus the airline’s current policy.
We will also talk about healthcare applications, in particular for the prevention of cardiovascular diseases. Deciding when to collect information, such as the patient’s cholesterol levels, is difficult. Measuring too frequently may be unnecessary and costly; on the other hand, measuring too infrequently means the patient may forgo needed treatment and experience adverse events related to the disease. We present results from estimating a stochastic model based on longitudinal data for cholesterol in a large cohort of patients seen in the national Veterans Affairs health system. We further use this model to study policies for when to collect measurements to assess the need for cholesterol-lowering medications.