Event Details

Making decisions when you do not trust your forecasts by Dr Pierre Pinson, Imperial College London

Abstract: Forecasts are most often there to be used as input to decision-making. And, since forecasts are never perfect, it appears natural to ask oneself how to make decisions given their uncertainty. A large body of scientific literature has developed within probabilistic forecasting and stochastic optimi-sation, in response to such challenges. We will focus on the specific case of newsvendor problems, which are common within procurement, logistics and energy trading for instance. However, we will extend the discussion to the case where one does not trust the probabilistic forecasts used as input, while also being unsure about the actual loss function at hand. After introducing the Bernouilli newsvendor problem, we present more general solutions to the newsvendor problem, when one does not trust the forecasts, for both decision variable and loss function parameters. A renewable energy trading example will allow illustrating the interest of the approach.