Measuring Accuracy between Ensemble Methods: AdaBoost.NC vs. SMOTE.ENN

The imbalanced class distribution is one of the main issue in data mining. This problem exists in multi class imbalance, when samples containing in one class are greater or lower than that of other classes. Most existing imbalance learning techniques are only designed and tested for two-class scenarios. The new negative correlation learning (NCL) algorithm for classification ensembles, called AdaBoost. NC, produce smaller error correlation along with training epochs. In existing system, an AdaBoost.NC is proposed to handle multiclass imbalance problems that fall into two major categories: data sampling and algorithmic modification. Cost–sensitive learning solutions combining both the data and algorithm level approaches assume higher misclassification costs with samples in the minority class and seek to minimize high cost errors. In this system, the main objective is to analyze the performance of data level proposals against algorithm level proposals, focusing on cost-sensitive models and thereby proposing a hybrid procedure that combines these two approaches. In order to analyze the over-sampling and under-sampling methodologies against Cost-sensitive learning approaches the ―synthetic minority oversampling technique‖ (SMOTE) is used. Keywords— Ensemble learning, SMOTE-ENN, Multi class imbalance problem, NCL.