Abstract:
Multidimensional ranking is useful to practitioners in political science,
computer science, social science, medical science, and allied fields. The
objective is to identify a consensus ranking of n objects that best fi ts independent rankings given by k different judges. The Kemeny distance is
used as a metric to obtain consensus ranking. For large n, under present
computing powers, it is not feasible to identify a consensus ranking. To
address the problem, researchers have proposed several algorithms. These
algorithms are able to handle datasets with n up to 200 in a reasonable
amount of time. However, run-time increases very quickly as n increases. In
the present paper, we propose two basic algorithms - Subiterative Convergence and Greedy Algorithm. Using these basic algorithms, two advanced
algorithms - FUR and SIgFUR are developed. We show that our results are
superior both in terms of Kemeny distance, as a performance measure, and
run-time to existing algorithms. The proposed algorithms, even for large n,
runs in few minutes.
