This package applies a resampling method to estimate rotated matrix for dimensional reduction. By applying the procedure, a number of new dimensions can be used as feature candidates for predictive modeling; thus, the number of candidates can be optimized depending on the training sample size. This helps to fulfill minimum events per variable (EPV) for a machine learning algorithm while optimizing the proportion of variance explained (PVE). Unlike other packages for dimensional reduction, this package applies resampling methods to prevent overfitting. The PVE optimization takes the predicted outcome into account without using it to represent the features.

Download R script

Reference manual