Meta-learning and Multi-objective Optimization to Design Ensemble of Classifiers

Ensemble of classifiers, or simply ensemble systems, have been proved to be efficient for pattern recognition tasks. However, its design can become a difficult task. For instance, the choice of its individual classifiers and the use of feature selection methods are very difficult to define in the design of these systems. In order to smooth out this problem, we will apply meta-learning and multi-objective optimization in the choice of important parameters of ensemble systems. Therefore, this work applies meta-learning techniques to define an initial configuration of an multi-objective optimization algorithm, more specifically NSGA II. The meta-learner is used to recommend the proportion of each type of base classifiers to compose the ensemble systems. The NSGA II is used to generate heterogeneous ensembles selecting attributes, types and parameters of base classifiers optimizing the classification error and the bad diversity. The results are analysed using error rate and multi-objective metrics in order to verify whether to use of meta-learning generates more accurate ensemble systems.