Designing an ensemble of classifiers is one of the popular research topics in machine learning since it can give better results than using constitute member. Furthermore, the performance of ensemble can be improved using the selection or adaptation approach. In the former, the optimal set of base classifiers, meta-classifier, original features, or meta-data is selected to obtain a better ensemble than using the entire classifiers and features. In the latter, base classifiers or combining algorithms working on the outputs of base classifiers are made to adapt to a particular problem. The adaptation here means that the parameters of these algorithms are trained to be optimal for each problem. In this study, we propose a novel evolving combining algorithm using the adaptation approach for the ensemble systems. Instead of using the numerical value when computing the representation for each class label, we propose to use the interval-based representation for the class label. The optimal value of the representation is found through Particle Swarm Optimization. The class label is assigned to each test instance by selecting the class label associated with the shortest distance between the predictions of the base classifiers on that instance and the interval-based representation. Experiments conducted on a number of popular datasets confirm that the proposed method is better than the well-known combining algorithms for ensemble systems using the combining methods including Decision Template, Sum Rule, L2-loss Linear Support Vector Machine, and Multiple Layer Neural Network, and the selection methods for ensemble systems (GA-Meta-data, META-DES, and ACO).

No abstract