IDENTYFIKACJA KOMPONENTÓW DESTRUKCYJNYCH W MODELACH PREDYKCYJNYCH W PODEJŚCIU WIELOMODELOWYM
IDENTIFICATION OF DESTRUCTIVE COMPONENTS IN PREDICTIVE MODELS WITH A MULTI-MODEL APPROACH

Summary/Abstract: In this paper we present a method of identification of destructive components in predictive models. This method may be applied in case of a multi-model approach and uses algorithms of blind signal separation. The evaluation of the characteristics of individual components will be based on the proposed metrics for evaluating the variation or smoothness of signals. In order to confirm the effectiveness of the presented method, we will present a practical experiment in which the results of the forecast of short-term electricity consumption are improved. Keywords: prediction, blind separation, ensemble methods, Theta noise measure

• Details
• Contents