Optimization of ensemble classifier system based on multiple objectives genetic algorithm

This paper introduces a mechanism to learn optimal classifier combining algorithms for an ensemble system. By using a genetic algorithm approach that focuses on 3 objectives namely the number of correct classified observations, the number of selected features and the number of selected classifiers, optimal solution can be achieved after several interactions of crossover and mutation. We also employ the Ordered Weighted Averaging operator in which a weight vector is built by a Linear Decreasing (LD) function to find average values of outputs from combining algorithms. Experiments on 2 well-known UCI Machine Learning Repository datasets demonstrate benefits of our approach compared with other state-of-the-art ensemble methods like Decision Template, SCANN and all fixed combining algorithms in the ensemble system.