Comparative Study of Pair wise learning in Imbalanced Data Problem Using Different Classification Techniques
Author: Fatima Shakeel
Category: Information Technology Management
Abstract:
Classification and prediction of rare cases in the patients is a very important task. In such cases the data sets are mostly imbalanced because the number of majority class (negative) outnumbers the minority class (positive) by a large proportion. There is an uneven distribution between the two classes. A lot of research has been conducted on binary class imbalanced data where one class (major) outnumbers the other class (minor) by a huge marginal difference. This research work focuses towards the classification of a data where there are multiple minor classes. Two approaches have been compared using ensemble methods like boosting and bagging and check the overall accuracy of minority classes.
Keywords: Bagging, AdaBoost, One versus One (OVO), One versus All (OVA)