RT info:eu-repo/semantics/article
T1 Non-removal strategy for outliers in predictive models: The PAELLA algorithm case
T2 Estrategia sin remoción para espureos en modelos predictivos: El caso del algoritmo PAELLA
A1 Castejón Limas, Manuel
A1 Alaiz Moretón, Héctor
A1 Fernández Robles, Laura
A1 Alfonso Cendón, Javier
A1 Fernández Llamas, Camino
A1 Sánchez González, Lidia
A1 Pérez García, Hilde
A2 Proyectos de Ingenieria
K1 Ingenierías
K1 Outliers
K1 Espurios
K1 1203.02 Lenguajes Algorítmicos
K1 1203
AB [EN] This paper reports the experience of using the PAELLA algorithm as a helper tool inrobust regression instead of as originally intended for outlier identification and removal.This novel usage of the algorithm takes advantage of the occurrence vector calculated bythe algorithm in order to strengthen the effect of the more reliable samples and lessen theimpact of those that otherwise would be considered outliers. Following that aim, a seriesof experiments are conducted in order to learn how to better use the information contained in the occurrence vector. Using a contrively difficult artificial dataset, a referencepredictive model is fit using the whole raw dataset. The second experiment reports theresults of fitting a similar predictive model but discarding the samples marked as outliersby PAELLA. The third experiment uses the occurrence vector provided by PAELLA inorder to classify the observations in multiple bins and fit every possible model changingwhich bins are considered for fitting and which are discarded in that particular model. Thefourth experiment introduces a sampling process before fitting in which the occurrencevector represents the likelihood of being considered in the training dataset. The fifth experimentconsiders the sampling process as an internal step to be performed interleavedbetween the training epochs. The last experiment compares our approach using weightedneural networks to a state of the art method.
PB Oxford University Press
SN 1367-0751
LK https://hdl.handle.net/10612/18266
UL https://hdl.handle.net/10612/18266
NO Castejón-Limas, M., Alaiz-Moreton, H., Fernández-Robles, L., Alfonso-Cendón, J., Fernández-Llamas, C., Sánchez-González, L., & Pérez, H. (2020). Non-removal strategy for outliers in predictive models: The PAELLA algorithm case. Logic Journal of the IGPL, 28(4), 418-429.https://doi.org/10.1093/jigpal/jzz052
DS BULERIA. Repositorio Institucional de la Universidad de León
RD 23-may-2024