Can I read Spatial-SMOTE for handling imbalance in spatial regression tasks on EtoBox?

Spatial-SMOTE for handling imbalance in spatial regression tasks by Rahul Dasharath Gavas; Monidipa Das; Soumya Kanti Ghosh; Arpan Pal is a Computer Science article available to read on EtoBox.

What is Spatial-SMOTE for handling imbalance in spatial regression tasks about?

Class imbalance in classification tasks leads to biased learning in the model training phase. In case of regression, the presence of rare cases over the normal cases in the target variable, also culminates in data imbalance. Since, the target variable is continuous in nature, the handling of imbalance in regression tasks is not straightforward when compared to that of classification tasks. Synthetic minority over-sampling technique (SMOTE) and its variants have been explored in upsampling rare events in imbalanced regression. These approaches solely rely on amplitude values of features and their nearest neighbours during the upsampling process. However, geo-tagged spatial data comes with an additional location information, i.e. latitude and longitude. This information needs to be incorporated during upsampling process in order to retain the spatial dependencies in the underlying data. We thus propose a new pipeline named, 'Spatial-SMOTE for regression', which deals with upsampling the rare events by retaining the significance of the spatial distribution of data. The effectiveness of this process is shown on two geo-tagged regression datasets for predicting the normal and the rare c

Who reads Spatial-SMOTE for handling imbalance in spatial regression tasks?

Author

Rahul Dasharath Gavas; Monidipa Das; Soumya Kanti Ghosh; Arpan Pal

Publisher

Springer Science and Business Media LLC

Published

2023

Language

EN

Field

Computer Science (Physical Sciences)