An Application of Oversampling, Undersampling, Bagging and Boosting in Handling Imbalanced Datasets

Most classifiers work well when the class distribution in the response variable of the dataset is well balanced. Problems arise when the dataset is imbalanced. This paper applied four methods: Oversampling, Undersampling, Bagging and Boosting in handling imbalanced datasets. The cardiac surgery dataset has a binary response variable (1 = Died, 0 = Alive). The sample size is 4976 cases with 4.2 % (Died) and 95.8 % (Alive) cases. CART, C5 and CHAID were chosen as the classifiers. In classification problems, the accuracy rate of the predictive model is not an appropriate measure when there is imbalanced problem due to the fact that it will be biased towards the majority class. Thus, the performance of the classifier is measured using sensitivity and precision Oversampling and undersampling are found to work well in improving the classification for the imbalanced dataset using decision tree. Meanwhile, boosting and bagging did not improve the Decision Tree performance.

[1]  Stan Matwin,et al.  Classifying Severely Imbalanced Data , 2011, Canadian Conference on AI.

[2]  Christophe Mues,et al.  An experimental comparison of classification algorithms for imbalanced credit scoring data sets , 2012, Expert Syst. Appl..

[3]  Nitesh V. Chawla,et al.  Data Mining for Imbalanced Datasets: An Overview , 2005, The Data Mining and Knowledge Discovery Handbook.

[4]  Gary M. Weiss.  Mining with rarity: a unifying framework , 2004, SKDD.

[5]  Dong-Sheng Cao,et al.  The boosting: A new idea of building models , 2010 .

[6]  Robert C. Holte,et al.  Severe Class Imbalance: Why Better Algorithms Aren't the Answer , 2005, ECML.

[7]  M. Mostafizur Rahman,et al.  Addressing the Class Imbalance Problem in Medical Datasets , 2013 .

[8]  Dimitris Kanellopoulos,et al.  Handling imbalanced datasets: A review , 2006 .

[9]  Taghi M. Khoshgoftaar,et al.  Comparing Boosting and Bagging Techniques With Noisy and Imbalanced Data , 2011, IEEE Transactions on Systems, Man, and Cybernetics - Part A: Systems and Humans.

[10]  Florentino Fernández Riverola,et al.  Evaluating the effect of unbalanced data in biomedical document classification , 2011, J. Integr. Bioinform..

[11]  Panayiotis E. Pintelas,et al.  Combining Bagging and Boosting , 2007 .

[12]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[13]  Antoine Geissbühler,et al.  Learning from imbalanced data in surveillance of nosocomial infection , 2006, Artif. Intell. Medicine.

[14]  Robert C. Holte,et al.  C4.5, Class Imbalance, and Cost Sensitivity: Why Under-Sampling beats Over-Sampling , 2003 .

[15]  S. Tom Au,et al.  Mining Rare Events Data by Sampling and Boosting: A Case Study , 2010, ICISTM.

[16]  Nitesh V. Chawla,et al.  SMOTE: Synthetic Minority Over-sampling Technique , 2002, J. Artif. Intell. Res..

[17]  Longbing Cao,et al.  Effective detection of sophisticated online banking fraud on extremely imbalanced data , 2012, World Wide Web.

[18]  Francisco Herrera,et al.  A Review on Ensembles for the Class Imbalance Problem: Bagging-, Boosting-, and Hybrid-Based Approaches , 2012, IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews).

[19]  David A. Cieslak,et al.  Automatically countering imbalance and its empirical relationship to cost , 2008, Data Mining and Knowledge Discovery.

[20]  Gustavo E. A. P. A. Batista,et al.  A study of the behavior of several methods for balancing machine learning training data , 2004, SKDD.

[21]  Ekrem Duman,et al.  Comparing alternative classifiers for database marketing: The case of imbalanced datasets , 2012, Expert Syst. Appl..