PCA $($or any other unsupervised dimensionality reduction$)$ can be used to transform data before fitting a supervised learning model. In this problem, we will apply PCA for dimensionality reduction, then use a support vector classifier fitted on a subset of the PCA$-$transformed features.

We will use K$-$fold cross validation to decide the number of principal components to use, according to the one$-$SE rule.

In this workspace, write code to split the data into training and test sets, and perform the analysis described above on the training set.

Name Type Description

acc$\_$mean $1$d numpy array Mean validation accuracy for each candidate model.

acc$\_$se $1$d numpy array Standard error of the mean of validation accuracy for each candidate model.

n$\_$pca$\_$opt integer Optimal number of components according to 'best mean validation accuracy' rule.

n$\_$pca$\_$one$\_$se integer Optimal number of components according to 'one SE$\text{'}$ rule.