This article discusses a novel Procrustes cross-validation method for augmenting numeric and mixed datasets, focusing on a mathematical framework for generating predictive value sets. It introduces two implementations based on Singular Value Decomposition and Partial Least Squares decomposition, essential for analyzing datasets where predictors and responses differ. The examples provided demonstrate applying Artificial Neural Networks (ANN) for regression and classification tasks with the Tecator and Heart datasets. This method aims to improve dataset understanding and predictive modeling by adopting a robust and structured approach to data augmentation.
To augment the data, a specific model should be developed that captures the variance-covariance structure of predictors, directly related to the responses.
The Procrustes cross-validation approach proposed in this study serves as a significant tool to enhance the analysis of numeric and mixed datasets.
#data-augmentation #procrustes-cross-validation #statistical-methods #machine-learning #predictive-modeling
Collection
[
|
...
]