Feature requests

[serenalotreck]

@peipeiwang6 wrote a separate ML pipeline with several new features, which should be incorporated into the lab's pipeline. The features are:

• Upsampling instead of downsampling for training and test sets
• Use permutation importance rather than gini importance when doing feature selection
• Perform stratified sampling for both test/train sets and during cross-validation
• Allow user to specify a number of samples to choose for balanced set, as opposed to a percentage