Non-Discriminating Criteria
In MCDA a criterion is non-discriminating if for all alternatives we have the same evaluation
For example,
| $f_1$ | $f_2$ | … | $f_k$ | … | $f_q$ | $a_1$ | $f_1(a_1)$ | $f_2(a_1)$ | … | $\alpha$ | … | $f_q(a_1)$ | $a_2$ | $f_1(a_2)$ | $f_2(a_2)$ | … | $\alpha$ | … | $f_q(a_2)$ | … | … | … | … | … | … | … | $a_n$ | $f_1(a_n)$ | $f_2(a_n)$ | … | $\alpha$ | … | $f_q(a_n)$ |
$f_k$ is non-discriminating criterion because for all alternatives it evaluates to $\alpha$
- this criteria is meaningless
- we want to delete it
- but we must be careful: because it may lead to Rank Reversal for some MCDA methods
- for PROMETHEE it will not: we can safely remove such criteria