Improving out-of-time predictive performance in business analytics via robust optimization and profit metrics
FONDO
Fondecyt regular N 1200221
AÑO VIGENCIA
2019-2023
Investigador principal
Sebastián Maldonado
Otros investigadores
Carla Vairetti, Juan Perez, Julio Lopez
Resumen
The proposal considers the creation of machine learning models that are robust in the presence of small changes in the data distribution. These changes can be anticipated either by modeling the trends and seasonal patterns present in the training set or by using a robust optimization approach, in which a pessimistic framework for the data distribution is assumed. On the other hand, the project seeks to adapt and customize predictive models that are robust in changing environments to business analytics applications.
The proposal considers the creation of machine learning models that are robust in the presence of small changes in the data distribution. These changes can be anticipated either by modeling the trends and seasonal patterns present in the training set or by using a robust optimization approach, in which a pessimistic framework for the data distribution is assumed. On the other hand, the project seeks to adapt and customize predictive models that are robust in changing environments to business analytics applications.