This article discusses the development and validation of a novel data augmentation method using PV-sets based on Singular Value Decomposition (SVD) and Partial Least Squares (PLS) decomposition techniques. Focused on datasets with moderate to high collinearity, the study demonstrates its application using ANN regression and classification on two datasets, including the Tecator meat sample dataset. Results indicate that the proposed augmentation improves model performance significantly. The methodology is implemented in Python and R, with scripts provided for the reproducibility of results, promoting accessibility in data analysis.
The study showcases the development of a novel data augmentation method using PV-sets derived from Singular Value Decomposition (SVD) and Partial Least Squares (PLS) methods to enhance datasets with high collinearity.
Results indicate that augmentation using PV-sets significantly improves both the regression and classification performance of artificial neural networks on the selected datasets, highlighting the method's effectiveness.
The implementation of this method was conducted using Python and R, making it accessible, as the Python scripts for reproducing the results are publicly available.
The Tecator dataset, used in the study, contains spectral measurements of meat samples, illustrating the versatility of the proposed method in various applications.
Collection
[
|
...
]