An empirical study of the noise impact on cost-sensitive learning
作者： Taghi M. Khoshgoftaar , Yong Shi , Xindong Wu , Xingquan Zhu
DOI:
关键词:
摘要: In this paper, we perform an empirical study of the impact noise on cost-sensitive (CS) learning, through observations how a CS learner reacts to mislabeled training examples in terms misclassification cost and classification accuracy. Our results theoretical analysis indicate that can raise serious concerns for classification, especially when misclassifying some classes becomes extremely expensive. Compared general inductive problem handling data cleansing is more crucial, should be carefully investigated ensure success learning.
ijcai.org参考文章(16)
Nada Lavrac, Ciril Groselj, Dragan Gamberger, Experiments with Noise Filtering in a Medical Domain international conference on machine learning. pp. 143- 151 ,(1999)
Peter D. Turney, Cost-sensitive learning bibliography the web conference. ,(2000)
Jeffrey P. Bradford, Clayton Kunz, Ron Kohavi, Cliff Brunk, Carla E. Brodley, Pruning decision trees with misclassification costs european conference on machine learning. pp. 131- 136 ,(1998) , 10.1007/BFB0026682
Ulf Brefeld, Peter Geibel, Fritz Wysotzki, Support vector machines with example dependent costs european conference on machine learning. pp. 23- 34 ,(2003) , 10.1007/978-3-540-39857-8_5
Xingquan Zhu, Xindong Wu, Cost-guided class noise handling for effective cost-sensitive learning international conference on data mining. pp. 297- 304 ,(2004) , 10.1109/ICDM.2004.10108
Michael Pazzani, Christopher Merz, Patrick Murphy, Kamal Ali, Timothy Hume, Clifford Brunk, Reducing Misclassification Costs Machine Learning Proceedings 1994. pp. 217- 225 ,(1994) , 10.1016/B978-1-55860-335-6.50034-9
Carla E. Brodley, Mark A. Friedl, Identifying mislabeled training data Journal of Artificial Intelligence Research. ,vol. 11, pp. 131- 167 ,(1999) , 10.1613/JAIR.606
Thomas G. Dietterich, Valentina Bayer Zubek, Pruning Improves Heuristic Search for Cost-Sensitive Learning international conference on machine learning. pp. 19- 26 ,(2002)
Pedro Domingos, MetaCost: a general method for making classifiers cost-sensitive knowledge discovery and data mining. pp. 155- 164 ,(1999) , 10.1145/312129.312220
Naoki Abe, Bianca Zadrozny, John Langford, An iterative method for multi-class cost-sensitive learning knowledge discovery and data mining. pp. 3- 11 ,(2004) , 10.1145/1014052.1014056