Improved ensemble learning for classification techniques based on majority voting

This paper proposes the methodology for improving the performance of the classification model, over several methods. The accuracy values obtained through experiments permit the evaluation of each method's performance. We propose a concept that brings Ensemble learning to model classification, in order to improve performance through majority voting, called M-Ensemble learning. The improved Ensemble learning approach is divided into two main formats of combined methods, namely the 3-Ensemble model (combining odd number methods, such as Naïve Bayes, Decision Tree, and Multilayer Perceptron); and the 4-Ensemble model (combining even number methods, such as Naïve Bayes, Decision Tree, Multilayer Perceptron, and K-Nearest Neighbor). The most improved classification model resulted from the improved 3-Ensemble method, with an accuracy value of 83.13%, compared with the Multilayer Perceptron based model classification and the 4-Ensemble model, which yielded accuracy values of 80.67% and 81.86%, respectively.