The Influence of Class Imbalance on Cost-Sensitive Learning: An Empirical Study

By Xu-Ying Liu, Zhi-Hua Zhou. 2006

280 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Electrical and Electronic Engineering

Citations in the corpus

Box drawings for learning with imbalanced data (2014, methodology): “We use a cost-sensitive learning approach in our methods, similar to Liu and Zhou (Liu and Zhou 2006) and McCarthy et al .”
Learning from Imbalanced Data (2009, background): “domains [11], [52], [53], cost-sensitive learning is superior to”
Cybercrime detection in online communications: The experimental case of cyberbullying detection in the Twitter network (2016, background): “Several approaches for overcoming these problems have been proposed, such as a combination of oversampling the minority (abnormal) class and under-sampling the majority (normal) class (Chawla, Bowyer, Hall, & Kegelmeyer, 2002), as well as weight adjusting approaches (Liu & Zhou, 2006).”
Dynamically Weighted Balanced Loss: Class Imbalanced Learning and Confidence Calibration of Deep Neural Networks (2021, methodology): “These methods have been reported to perform well on binary data [32].”
Cost-sensitive learning methods for imbalanced data (2010, methodology): “• The second method for CSL, instead of assuming that we know the cost ratio (or cost matrix) before learning as in the first method and other previous works ([30], [21], [25]) or setting the cost ratio by inverting of prior class distributions ([31], [7], [32]), we treat this number as a hyperparam”
Disentangling Label Distribution for Long-tailed Visual Recognition (2020, methodology): “Most long-tailed visual recognition methods can be divided into two strategies: modifying the data sampler to balance the class frequency during optimization [7, 21, 25, 8, 51], and modifying the class-wise weights of the classification loss to increase the importance of tail classes in terms of emp”
Financial credit risk assessment: a recent review (2015, methodology): “To manipulate the two problems simultaneously, a possible solution is to merge the class imbalance ratio with cost ratio through rescaling (Liu and Zhou 2006).”