A novel ensemble classifier based on multiple diverse classification methods

Classification is one of the most important tasks in machine learning. The ensemble classifier which consists of a number of basic classifiers is an efficient classification technique and has shown its effectiveness in many applications. The diversity and strength of the basic ones are two main elements which influence the performance of the ensemble classifier. Since different classification methods could capture the different discriminative information of the data by different classification criteria, using different classification techniques to build the basic ones could increase their diversity and strength. This paper proposes a new ensemble learning method which combines three different learning techniques to build the ensemble basic learners and adopts a double-layer voting method to enhance the strength and diversity of the basic ones, simultaneously. The new method is tested on six benchmark datasets from UCI machine learning repository. The experimental results show that the proposed method outperforms the other ensemble techniques and single classifiers in the classification accuracy in most cases.