With the advent of decision science, significant elucidation has been sought in the literature of multi criteria decision making. Often, it is observed that for the same MCDM problem, different methods fetch way-apart ranks and the phenomenon leads to rank reversal. To alleviate this problem, different methodologies like the Borda rule, the Copeland method, the Condorcet method, the statistical Thurstone scaling, and linear programming methods are readily available in the literature. In connection with the same, the authors proposed a novel technique to aggregate the ranks laid by different methods. The algorithm initially assigns equal weights to the methods involved to avoid biasness to a particular method and a simple average rank was obtained. Then, after the separation measures of individual methods with respect to average rank were calculated. Considering the separation measure the higher the weightage, the dynamic weights are ascertained to declare the weighted aggregate rank subjected to the terminal condition which include whether the previous rank equals to the current rank or not. To substantiate the proposed algorithm, a materials selection problem was taken into consideration and solved with the proposed technique. Moreover, the same problem was solved by existing voting techniques like the Borda and the Copeland-Condoract methods. The authors found a correlation of more than 85% between the proposed and existing methodologies.
A Genetic Algorithm Approach for Group Recommender System Based on Partial Rankings
