Hybrid Pareto Differential Evolutionary Artificial Neural Networks to Determined Growth Multi-classes in Predictive Microbiology

The main objective of this work is to automatically design artificial neural network, ANN, models with sigmoid basis units for multiclassification tasks in predictive microbiology. The classifiers obtained achieve a double objective: high classification level in the dataset and high classification level for each class. For learning, the structure and weights of the ANN we present an Hybrid Pareto Differential Evolution Neural Network (HPDENN), a Differential Evolutionary approach based on the PDE multiobjective evolutionary algorithm. The PDE algorithm is augmented with a local search using the improved Resilient Backpropagation with backtraking-IRprop+ algorithm. To analyze the robustness of this methodology, we have applied it to two complex problems of classification in predictive microbiology (Staphylococcus Aureus and Shigella Flexneri). The results obtained in Correct Classification Rate (C) and Minimum Sensitivity (S) for each class show that the generalization ability and the classification rate in each class can be more efficiently improved within this multiobjective algorithm.