Bad and good errors: value-weighted skill scores in deep ensemble learning
Abstract
In this paper we propose a novel approach to realize forecast verification. Specifically, we introduce a strategy for assessing the severity of forecast errors based on the evidence that, on the one hand, a false alarm just anticipating an occurring event is better than one in the middle of consecutive non-occurring events, and that, on the other hand, a miss of an isolated event has a worse impact than a miss of a single event, which is part of several consecutive occurrences. Relying on this idea, we introduce a novel definition of confusion matrix and skill scores giving greater importance to the value of the prediction rather than to its quality. Then, we introduce a deep ensemble learning procedure for binary classification, in which the probabilistic outcomes of a neural network are clustered via optimization of these value-weighted skill scores. We finally show the performances of this approach in the case of three applications concerned with pollution, space weather and stock prize forecasting.
Keywords
Cite
@article{arxiv.2103.02881,
title = {Bad and good errors: value-weighted skill scores in deep ensemble learning},
author = {Sabrina Guastavino and Michele Piana and Federico Benvenuto},
journal= {arXiv preprint arXiv:2103.02881},
year = {2021}
}