Improving classification of imbalanced student dataset using ensemble method of voting, bagging, and adaboost with under-sampling technique

Abstract

Student imbalanced data is one of the problems in data mining community. To state the student dropout problem, an ensemble method with under-sampling technique is applied for improved the performance of classification of imbalanced student dataset. Mutual information for feature selection methods is used to find a significant feature. Voting, bagging, and adaboost technique in the ensemble method are used with decision tree (C4.5) and artificial neural network (ANN) classifiers to classify student in point of research objective. The result of this experiment evaluated by overall accuracy, precision, and recall. Bagging technique by random forest gave the best result in terms of overall accuracy is 74.57% and the recall of the prediction in the class (low) which we interested is 95.61%. This experiment extremely useful not only finding a useful knowledge for student and academic planning and management but also improving classification for imbalanced data which is the most effective way to state the classify student performance.