Sample Size on the Impact of Imbalance Learning

Abstract:

Classification of imbalanced data sets is widely used in many real life applications. Most state-of-the-art classification methods which assume the data sets are relatively balanced lose their efficiency. The paper discusses the factors which influence the modeling of a capable classifier in identifying rare events, especially for the factor of sample size. Carefully designed experiments using Rotation Forest as base classifier, carried on 3 datasets from UCI Machine Learning Repository based on weak show that, in particular imbalance ratio, increases the size of training set by unsupervised resample the large error rate caused by the imbalanced class distribution decreases. The common classification algorithm can reach good effect.

• Periodical: Advanced Materials Research (Volumes 756-759)
• Pages: 2547-2551
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.756-759.2547
• Online since: September 2013
• Authors: Wei Mei Zhi, Hua Ping Guo, Ming Fan
• Keywords: Classification, Imbalanced Data Sets, Principal Component Analysis (PCA), Sampling Size

References

[1] P-N. Tan and M. Steinbach, Introduction to Data Mining, p.127–187, ［M］ (2005).

[2] Fawcett, T. and Provost, F., Combining Data Mining and Machine Learning for Effective User Profile, Proceedings of the 2nd International Conference on Knowledge Discovery and Data Mining, Protland OR, AAAI Press(1996) 8-13.

[3] Lewis, D. and Caltett, Heterogeneous, Uncertainty Sampling for Supervized Learning. Proceedings of the 11th International Conference on Machine Learning. ICML'94(1994)148-156.

[4] Y. Sun, M. S. Kamel and A. K.C. Wong, Cost-sensitive boosting for classification of imbalanced data, Patter Recognition Society, pp.3358-3378 , (2007).

DOI: 10.1016/j.patcog.2007.04.009

[5] S. Visa and A. Ralescu, Issues in Mining imbalanced Data Sets-A Review Paper, Proc. Of MidWest Artificial Intelligence and Cognitive Science Conference, pp.67-73, (2005).

[6] G. Weiss and F. Provost, Learning when training data are costly: the effect of class distribution on tree induction, J. Aritif. Intell. Res. 19, pp.315-354 , (2003).

DOI: 10.1613/jair.1199

[7] K. Ezawa, M. Singh and S. W. Norton, Learning goal oriented Bayesian networks for telecommunications risk management, in: Proceedings of the Thirteenth International Conference on Machine Learning, Bari, Italy, pp.139-147, (1996).

[8] N. V. Chawla, K. W. Bowyer, L. O. Hall, and W. P. Kegelmeyer, SMOTE: Synthetic Minority Over-Sampling Technique, J. Artificial Intelligence Research, vol. 16, pp.321-357, (2002).

DOI: 10.1613/jair.953

[9] R. Agarwal, and M. V. Joshi, PNrule: A new Framework for Learning Classifier Models in Data Mining (A Case-Study in Network Intrusion Detection), Technical Report TR 00-01, Department of Computer Science University of Minnesota, USA, (2000).

DOI: 10.1137/1.9781611972719.29

[10] Cieslak, D.A., CHawla, N.V., Learning decision trees for unbalanced data. ECML 99. 241-256(2008).