Ensemble Machine Learning for Multi-Class Classification Tasks

Ensemble learning is the process of amalgamating multiple classifiers to engender a strong classifier in supervised machine learning. It boost-up the performance week learning algorithms. In this paper, we have tested the performance of RandonForest, Bagging, and Boosting methods applying decision tree, Naïve Bayes classifier, and NBTree as a base classifier on 10 benchmark datasets. RandomForest and bagging methods use a majority voting technique, whereas boosting (AdaBoost) algorithm uses a weighted majority technique to classify new instances. A top-down recursive divide and conquer recursive algorithm is decision tree as well as naïve Bayes classifier work as a deterministic classifier. NBTree is a combination of a decision tree and naïve Bayes classifier where nodes of a decision tree are replaced by Bayes models. The experimental results demonstrate how flexible and powerful ensemble learning approaches are for dealing with classification problems. In order to help practitioners choose the best ensemble approach, several methods are methodically compared to show their respective strengths and disadvantages..