
362 ◾ Waad Bouaguel, Ghazi Bel Mufti, and Mohamed Limam
Solution to Disjoint Ranking for Similar Features
We take as staring point the ranked list Initial
R
, where each item r
of this list,
presents the rank of a feature f
i
. Remind that we always give 1 as optimal ranking
to the most signicative feature, rankings are always positive, and lower rank shows
higher preference in the list.
In each iteration, we study the similarity between the rst feature (i.e., the
feature with the most relevance to the target concept and where r
fi
=1) and
the remaining features in the aggregated list, and for that, we use the similarity
function (SIM). is function ...