A genetic distance metric to discriminate the selection of algorithms for the general ATSP problem

The only metric that had existed so far to determine the best algorithm for solving an general Asymmetric Traveling Salesman Problem (ATSP) instance is based on the number of cities; nevertheless, it is not sufficiently adequate for discriminating the best algorithm for solving an ATSP instance, thus the necessity for devising a new metric through the use of data-mining techniques. In this paper we propose: (1) the use of a genetic distance metric for improving the selection of the algorithms that best solve a given instance of the ATSP and (2) the use of discriminant analysis as a means for predictive learning (data-mining techniques) aiming at selecting meta-heuristic algorithms.