grouping correlated features together and replacing each cluster with a single representative or aggregate. This is particularly useful when you have many features measuring similar constructs (e.g., multiple metrics of defensive quality).