Feature selection allows you to remove irrelevant features from your dataset prior to the learning process. The
caret package provides several implementations of feature selection methods. Most of these implementations are supervised approaches, where you can include information about the outcome (class/response variable) as part of your selection criteria.
There are, however, two simple unsupervised feature selection strategies in
caret that could prove quite helpful down the road, i.e., (1) removing features with zero or near-zero variance and (2) removing highly correlated features. Before we explain how to do that, let's test your knowledge about the three different types of feature selection methods.
You implemented a method to explore all possible subsets of 5 features in your data and select the one that maximizes KNN performance. Which type of feature selection method is that one?