Hybrid ensembles of decision trees and Bayesian network for class imbalance problem

Class imbalance problem is the main issue causing unsatisfactory outcome in classification. Any type of classification used still cannot improve the result. Therefore, in this research we propose a new hybrid ensemble model based on AdaBoost.M2 and adopt SMOTE algorithm to solve the class imbalance problem in order to predict the probability of term deposit from bank customers. The proposed hybrid ensemble model consist of diverse based classifiers which are Bayesian Network, Alternating Decision Tree, Tree-J48, and REPTree (Reduced-Error Pruning). From the experimental results, the proposed model can achieve the highest performance comparing to normal ensemble models and ensemble models that use majority class reduction, and finally generates the results of 91.5% sensitivity, 100% specificity, and 96.3% accuracy.