Multiobjective Pareto Ordinal Classification for Predictive Microbiology

This paper proposes the use of a Memetic Multiobjective Evolutionary Algorithm (MOEA) based on Pareto dominance to solve two ordinal classification problems in predictive microbiology. Ordinal classification problems are those ones where there is order between the classes because of the nature of the problem. Ordinal classification algorithms may take advantage of this situation to improve its classification. To guide the MOEA, two non-cooperative metrics have been used for ordinal classification: the Average of the Mean Absolute Error, and the Maximum Mean Absolute Error of all the classes. The MOEA uses an ordinal regression model with Artificial Neural Networks to classify the growth classes of microorganisms such as Listeria monocytogenes and Staphylococcus aureus.