If A (density of a material) and B (solid volume of a material=mass/density) are highly correlated features, how to decide which one of these two features shall I drop?
After normalization, feature A consists of two values, either 0 or 1 (I have only used 2 materials in my model), but feature B consists of more than 5 values.
My guess is to keep feature B due to its high variance.
