A Multivariate Data Analysis Procedure for Measuring and Comparing the Quality of Performance of Heuristics

The search effort is a heuristic characteristic which determines the quality of the feasible solution. However, the feature has been overshadowed by the convergence and deviation rates as a result of more activities in development experiments than in scientific experiments. It is therefore, our concern that with the growing call for resuscitation of scientific experiments, the need to measure the quality of the feasible solutions will proportionally increase in the foreseeable future. Hence, we develop a multivariate free distribution data analysis procedure for measuring search effort rates, quality of best fits and search distributions of the heuristics. The data analysis procedure can be applied even if there are missing observations. We apply heuristics to solve the travelling salesperson problem in order to generate data for practical demonstration of the application of the new development. The evaluation of the performance of the data analysis procedure had so far produced satisfactory results.