Intelligent Resampling for Imbalanced Big Data Classication: Novel Methods and Applications
FONDO
Fondecyt Iniciación Nro 11200007
AÑO VIGENCIA
2020-2022
INVESTIGADORA
Carla Vairetti
Resumen
This proposal is designed under the hypothesis that the creation of synthetic instances of the minority class is not strictly necessary when millions of samples and hundreds of explanatory variables are available. However, random undersampling has some limitations when dealing with noise, and therefore researchers and practitioners have favored the use of oversampling over random undersampling even in big data environments. Therefore, this project proposes intelligent hybrid undersampling/oversampling strategies for efficient large-scale binary classification, overcoming the main shortcomings of random undersampling and intelligent oversampling.
Investigadora: Carla Vairetti
This proposal is designed under the hypothesis that the creation of synthetic instances of the minority class is not strictly necessary when millions of samples and hundreds of explanatory variables are available. However, random undersampling has some limitations when dealing with noise, and therefore researchers and practitioners have favored the use of oversampling over random undersampling even in big data environments. Therefore, this project proposes intelligent hybrid undersampling/oversampling strategies for efficient large-scale binary classification, overcoming the main shortcomings of random undersampling and intelligent oversampling.
Investigadora: Carla Vairetti