M-Bagging: A New Modified Bagging Classification Model to Improve Prediction accuracy

---

Abstract

Ensemble learning is a one kind of machine learning technique that improves the performance and robustness of the classification models and how the outputs of base classifiers are combined is one of the fundamental challenges in ensemble learning systems. Among different types of ensemble learning models, Bagging is most popular due to its simplicity but Bagging has several drawbacks. As for example in bootstrapped creation out of bag samples are not used properly or it does not take care for misclassified samples and it uses homogeneous classifiers. So, in this work, we have developed a modified bagging ensemble classification model by embedding modified bootstrapping techniques so that misclassified samples are specially taken care, out of bag samples are also taken care. Apart from these several heterogeneous classifiers are also used here in novel manner. From experimental results it has been found that the proposed model is superior compared to other existing basic ensemble models as well as other state of the art models.

---

Keywords

Ensemble learning; Bagging; Bootstrapping; Classification; Class imbalance.