Machine Learning from Imbalanced Data Sets 101

By Foster J. Provost. 2008

532 citations recorded by the source corpus.

How researchers see this paper

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

In Evaluation & Benchmarks

Citations in the corpus

Measuring Data (2022, background): “…extent to which the probability distribution of a variable deviates from the normal distribution) can help orient subsequent modeling approaches and indicate whether additional steps are necessary, for instance in terms of class balancing or data normalization (Prati et al., 2009; Provost, 2000).”
Learned lessons in credit card fraud detection from a practitioner perspective (2014, background): “In an unbalanced class problem, it is well-known that quantities like TPR ( TPTPþFN), TNR ( TN FPþTN) and Accuracy ( TPþTN TPþFNþFPþTN) are misleading assessment measures (Provost, 2000).”
Oversampling for Imbalanced Learning Based on K-Means and SMOTE (2017, methodology): “The optimal imbalance ratio is not obvious and has been discussed by other researchers (Provost, 2000; Estabrooks et al., 2004).”
Support vector machines in DSC‐based glioma imaging: Suggestions for optimal characterization (2010, methodology): “, use class weights on the decision function of the classifier (40).”
Classification with class imbalance problem: A review (2015, background): “The authors in the same publication also highlighted that computational cost is higher when more examples are replicated from a class, besides leading to over-fitting [17][61].”
The Precision-Recall Plot Is More Informative than the ROC Plot When Evaluating Binary Classifiers on Imbalanced Datasets (2015, background): “The classification of imbalanced datasets is a relatively new challenge in the field of machine learning [5, 10].”
ADASYN: Adaptive synthetic sampling approach for imbalanced learning (2008, background): “From applications in Web mining to text categorization to biomedical data analysis [1], this challenge manifests itself in two common forms: minority interests and rare instances.”
Learning on the border: active learning in imbalanced data classification (2007, background): “Despite that they are dif.cult to identify, rare instances generally constitute the target concept in classi.cation tasks.”
Learning from Imbalanced Data (2009, background): “Recently, however, various issues on active learning from imbalanced data sets have been discussed in literature [97], [98], [99], [100].”
Deep Long-Tailed Learning: A Survey (2021, background): “3) Logit Adjustment: Logit adjustment [14], [140] seeks to resolve the class imbalance by adjusting the prediction logits of a class-biased deep model.”
Imbalanced Deep Learning by Minority Class Incremental Rectification (2018, background): “In contrast, the threshold-adjustment technique changes the decision threshold in test time [24,51,52,53,54].”
Long-tail learning via logit adjustment (2020, methodology): “Our framework revisits the classic idea of logit adjustment based on label frequencies (Provost, 2000; Zhou & Liu, 2006; Collell et al., 2016), which encourages a large relative margin between a pair of rare positive and dominant negative labels.”
Adaptive Machine Learning for Credit Card Fraud Detection (2015, background): “4, standard classification measure such as MME, BER, TPR and TNR are misleading assessment measures in unbalanced class problem [49].”
Online and Offline Domain Adaptation for Reducing BCI Calibration Effort (2017, background): “Of course, there are many other approaches for handling clas s imbalance [18], [26], [37].”
Cost-sensitive boosting for classification of imbalanced data (2007, background): “However, standard classifiers generally perform poorly on the imbalanced data sets as they are designed to generalize from training data, and pay less attention to the rare cases [30, 53, 68, 76].”