Integration of predictive and prescriptive approaches in stochastic optimization
The use of data science tools for decision making in organizations has grown significantly in recent years. The usual practice is to first use the data to create a forecast model for uncertain elements, such as demand, and then use this model as input in an optimization scheme that chooses the best course of action from a set of alternatives. The objective of this project is to build prescriptive models that integrate these two steps, so that the impact of the forecast that is made from the data on the decision model can be directly measured. The project proposes to apply the methodology to problems in energy planning, an area in which data abound and in which optimization models under uncertainty have been widely used.