Terminology: what is the correct word for the quantity of "association" between x value and classes

I am looking for a word for the quantity of association between a variable x value and a classification class.

For example, let's imagine we have two two classes $A$ and $B$ and two inputs $\mathbf{x}_A$ and $\mathbf{x}_B$ from each class respectively. The probability of membership in each class varies roughly linearly between these two inputs. I want a word to say the association of the input $\mathbf{x}_C = a \mathbf{x}_A + b \mathbf{x}_B$ with $A$ is determined by $a$ and $b$.

I initially used the word correlation without considering it, but then I realised this is not correct.

Other usage examples:

• Feature $i$ has a negative association with class $A$
• We may add vector $\mathbf{u}$ because it is unassociated with the output classes.
• Transformation $f$ preserves association with the output class.
• Removing noise increases output class association.

I could say probability of membership, but that would make certain phrases awkward. For example Feature $i$ has a negative contribution to the probability of membership of class $A$. I would much rather say something similar to Feature $i$ has a negative correlation with class $A$.