An Ordinal Regression Approach for the Unequal Area Facility Layout Problem

This paper proposes the use of ordinal regression for helping the evaluation of Unequal Area Facility Layouts generated by an interactive genetic algorithm. Using this approach, a model obtained taking into account some objective factors and the subjective evaluation of the experts is constructed. Ordinal regression is used in this case because of the ordinal ranking between the different possible evaluations of the facility layouts made by the experts: {very deficient, deficient, intermediate, good, very good}. To do so, we will also make an approximation to some of the most successful ordinal classification methods in the machine learning literature. The best model obtained will be used in order to guide the searching of a genetic algorithm for generating new facility layouts.