Academic Performance Prediction Using Imbalance Classification Methods: A Study

Abstract

An issue of classification may arise when learning classifiers use skewed or imbalanced datasets. In case of an imbalanced dataset, the majority of examples are with one class whereas the other class which is normally considered as the most important class, is however signified by a minor share of instances. By using this type of data, the outcome of machine-learning models would be ineffective. There is a term (High training reliability) used to define preconceptions in between one instance against all other illustrations of the class. In this paper, most important methods used to solve the class imbalance problem i.e. data-level, algorithm-level, hybrid, cost-sensitive learning, deep learning etc. including their advantages and limitations are discussed in detail. Performance and efficiency of classifiers are evaluated using a numerous evaluation metrics such as Accuracy, Precision, Recall, F-Measure.