A Novel Ensemble Method for Imbalanced Data Classification

Abstract :- Classification algorithms have shown exceptional prediction results in the supervised learning area. These classification algorithms are not always efficient when it comes to real-life datasets due to class distributions. As a result, datasets for real-life applications are generally imbalanced. Numerous methods have been developed to treat imbalanced datasets, which can be divided into three categories: (1) data resampling (2) algorithm modification and (3) ensemble methods. Among those categories, ensemble methods are the important area that proves to improve the classification performance. Ensemble learning combines several base models, where a traditional algorithm is used to learn each of them. It aggregates the outputs from a set of different classifiers to correctly classify new data points. Some popular ensemble learning methods include Bagging, Boosting and Adaboost. Bagging is an inherent parallel ensemble learning technology whose components can be running at the same time, and uses majority voting or weighted majority voting to aggregate results. It has provided considerable performance gains over a single learner in many application domains. This paper proposed an ensemble methods using automobile data by fusing classifiers such as RBF and SVM with bagging and their performances are analyzed in terms of accuracy. A wide range of comparative experiments are conducted for standard dataset of automobile. The proposed bagged ensemble methods provide significant improvement of accuracy compared to individual classifiers and previous works on standard dataset of automobile are exhibited.