Label noises exist in many applications, which usually add difficulties for data analysis. A straightforward and effective method is to detect and filter out them prior to training. Ensemble learning based filter has shown promising performances. We define an important parameter to improve the performance of the algorithm. The proposed method is cost sensitive which integrates the mislabeled training dataset and noise costs for learning. Finally, the experimental results on the benchmark datasets show the superiority of the proposed method.
CITATION STYLE
Zhu, W., Yuan, H., Wang, L., Wan, M., Li, X., & Ren, J. (2019). A Novel Noise Filter Based on Multiple Voting. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 11632 LNCS, pp. 159–170). Springer Verlag. https://doi.org/10.1007/978-3-030-24274-9_14
Mendeley helps you to discover research relevant for your work.