ABSTRACT
Prediction bias is a well-known problem in classification algorithms, which tend to be skewed towards more represented classes. This phenomenon is even more remarkable in multi-label scenarios, where the number of underrepresented classes is usually larger. In light of this, we hereby present the Prediction Bias Coefficient (PBC), a novel measure that aims to assess the bias induced by label imbalance in multi-label classification. The approach leverages Spearman's rank correlation coefficient between the label frequencies and the F-scores obtained for each label individually. After describing the theoretical properties of the proposed indicator, we illustrate its behaviour on a classification task performed with state-of-the-art methods on two real-world datasets, and we compare it experimentally with other metrics described in the literature.
Supplemental Material
- Haibo He and Yunqian Ma. Imbalanced learning: foundations, algorithms, and applications. 2013. Google ScholarDigital Library
- Joffrey L Leevy, Taghi M Khoshgoftaar, Richard A Bauder, and Naeem Seliya. A survey on addressing high-class imbalance in big data. Journal of Big Data, 5(1):1--30, 2018.Google ScholarCross Ref
- Grigorios Tsoumakas and Ioannis Katakis. Multi-label classification: An overview. International Journal of Data Warehousing and Mining (IJDWM), 3(3):1--13, 2007.Google ScholarCross Ref
- Bin Liu, Konstantinos Blekas, and Grigorios Tsoumak. Multi-label sampling based on local label imbalance. arXiv preprint arXiv:2005.03240, 2020.Google Scholar
- Min-Ling Zhang, Yu-Kun Li, Hao Yang, and Xu-Ying Liu. Towards class-imbalance aware multi-label learning. IEEE Transactions on Cybernetics, 2020.Google ScholarCross Ref
- Yu Zhang, Yin Wang, Xu-Ying Liu, Siya Mi, and Min-Ling Zhang. Large-scale multi-label classification using unknown streaming images. Pattern Recognition, 99:107100, 2020.Google ScholarCross Ref
- Fangfang Luo, Wenzhong Guo, Yuanlong Yu, and Guolong Chen. A multi-label classification algorithm based on kernel extreme learning machine. Neurocomputing, 260:313--320, 2017.Google ScholarDigital Library
- Deborah Hellman. Measuring algorithmic fairness. Va. L. Rev., 106:811, 2020.Google Scholar
- Jon Kleinberg, Jens Ludwig, Sendhil Mullainathan, and Ashesh Rambachan. Algorithmic fairness. In Aea papers and proceedings, volume 108, pages 22--27, 2018.Google ScholarCross Ref
- Min-Ling Zhang and Zhi-Hua Zhou. A review on multi-label learning algorithms. IEEE transactions on knowledge and data engineering, 26(8):1819--1837, 2013.Google Scholar
- Haibo He and Edwardo A Garcia. Learning from imbalanced data. IEEE Transactions on knowledge and data engineering, 21(9):1263--1284, 2009. Google ScholarDigital Library
- Li Li and Houfeng Wang. Towards label imbalance in multi-label classification with many labels. arXiv preprint arXiv:1604.01304, 2016.Google Scholar
- Alberto Fernández, Victoria López, Mikel Galar, Mar'iA José Del Jesus, and Francisco Herrera. Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches. Knowledge-based systems, 42:97--110, 2013. Google ScholarDigital Library
- Zachary Daniels and Dimitris Metaxas. Addressing imbalance in multi-label classification using structured hellinger forests. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 31, 2017. Google ScholarDigital Library
- Thibaut Durand, Nazanin Mehrasa, and Greg Mori. Learning a deep convnet for multi-label classification with partial labels. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 647--657, 2019.Google ScholarCross Ref
- Francisco Charte, Antonio J Rivera, Mar'ia J del Jesus, and Francisco Herrera. Addressing imbalance in multilabel classification: Measures and random resampling algorithms. Neurocomputing, 163:3--16, 2015. Google ScholarDigital Library
- Muhammad Atif Tahir, Josef Kittler, and Fei Yan. Inverse random under sampling for class imbalance problem and its application to multi-label classification. Pattern Recognition, 45(10):3738--3750, 2012. Google ScholarDigital Library
- Francisco Charte, Antonio J Rivera, Mar'ia J del Jesus, and Francisco Herrera. Mlsmote: approaching imbalanced multilabel learning through synthetic instance generation. Knowledge-Based Systems, 89:385--397, 2015. Google ScholarDigital Library
- Jonathan Ortigosa-Hernández, Inaki Inza, and Jose A Lozano. Measuring the class-imbalance extent of multi-class problems. Pattern Recognition Letters, 98:32--38, 2017. Google ScholarDigital Library
- Ignazio Pillai, Giorgio Fumera, and Fabio Roli. Designing multi-label classifiers that maximize f measures: State of the art. Pattern Recognition, 61:394--404, 2017. Google ScholarDigital Library
- Francisco Charte, Antonio Rivera, Mar'ia José del Jesus, and Francisco Herrera. A first approach to deal with imbalance in multi-label datasets. In International Conference on Hybrid Artificial Intelligence Systems, pages 150--160. Springer, 2013.Google ScholarCross Ref
- Rui Zhu, Ziyu Wang, Zhanyu Ma, Guijin Wang, and Jing-Hao Xue. Lrid: A new metric of multi-class imbalance degree based on likelihood-ratio test. Pattern Recognition Letters, 116:36--42, 2018.Google ScholarCross Ref
- Marcus A Maloof. Learning when data sets are imbalanced and when costs are unequal and unknown. In ICML-2003 workshop on learning from imbalanced data sets II, volume 2, pages 2--1, 2003.Google Scholar
- Krzysztof Dembczynski, Arkadiusz Jachnik, Wojciech Kotlowski, Willem Waegeman, and Eyke Hüllermeier. Optimizing the f-measure in multi-label classification: Plug-in rule approach versus structured loss minimization. In International conference on machine learning, pages 1130--1138. PMLR, 2013. Google ScholarDigital Library
- Charles Spearman. The proof and measurement of association between two things. 1961.Google ScholarCross Ref
- Jerrold H Zar. Spearman rank correlation. Encyclopedia of biostatistics, 7, 2005.Google Scholar
- Joost CF de Winter, Samuel D Gosling, and Jeff Potter. Comparing the pearson and spearman correlation coefficients across distributions and sample sizes: A tutorial using simulations and empirical data. Psychological methods, 21(3):273, 2016.Google ScholarCross Ref
- Douglas G Bonett and Thomas A Wright. Sample size requirements for estimating pearson, kendall and spearman correlations. Psychometrika, 65(1):23--28, 2000.Google ScholarCross Ref
- Mohamed Aly. Survey on multiclass classification methods. Neural Netw, 19:1--9, 2005.Google Scholar
- Keiron O'Shea and Ryan Nash. An introduction to convolutional neural networks. arXiv preprint arXiv:1511.08458, 2015.Google Scholar
- Siwei Lai, Liheng Xu, Kang Liu, and Jun Zhao. Recurrent convolutional neural networks for text classification. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 29, 2015. Google ScholarDigital Library
- Marcin Michał Miro'nczuk and Jarosław Protasiewicz. A recent overview of the state-of-the-art elements of text classification. Expert Systems with Applications, 106:36--54, 2018.Google ScholarCross Ref
- Lawrence Mosley. A balanced approach to the multi-class imbalance problem. 2013.Google ScholarCross Ref
Index Terms
- Evaluating the Prediction Bias Induced by Label Imbalance in Multi-label Classification
Recommendations
Correlated Multi-label Classification with Incomplete Label Space and Class Imbalance
Special Section on Advances in Causal Discovery and Inference and Regular PapersMulti-label classification is defined as the problem of identifying the multiple labels or categories of new observations based on labeled training data. Multi-labeled data has several challenges, including class imbalance, label correlation, incomplete ...
Multi-label classification by exploiting label correlations
Nowadays, multi-label classification methods are of increasing interest in the areas such as text categorization, image annotation and protein function classification. Due to the correlation among the labels, traditional single-label classification ...
Learning label-specific features with global and local label correlation for multi-label classification
AbstractMulti-label algorithms often use an identical feature space to build classification models for all labels. However, labels generally express different semantic information and should have their own characteristics. A few algorithms have been ...
Comments