Learning and making decisions when costs and probabilities are both unknown

By Bianca Zadrozny, Charles Peter Elkan. Knowledge Discovery and Data Mining, 2001

537 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Information Systems

View Paper: https://www.semanticscholar.org/paper/d33c6d671580bc5104db395c4004370c7c299db3

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 Information Systems

Citations in the corpus

Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches (2013, background): “For example, solutions at data level [5,15,35] suffer from the increased search space, and solutions at algorithm level [66,31] become more complicated, as the learning algorithm must consider several small classes.”
Imbalanced Deep Learning by Minority Class Incremental Rectification (2018, methodology): “(2) Algorithmlevel: Modifying existing algorithms to give more emphasis on the minority classes [8,9,10,43,44,45,46,47,48,49,50].”
Mining with rarity: a unifying framework (2004, background): “Accuracy may be used to determine the class label, but this will yield an unreliable estimate when the number of examples that are covered is small [39; 54].”
Financial credit risk assessment: a recent review (2015, methodology): “The algorithm-level methods aim to strengthen the learning performance of existing algorithms with regards to the small class using some inductive bias, for example new pruning strategies of DTs (Zadrozny and Elkan 2001) and different penalty constants of SVM (Lin et al. 2002).”
Cost-sensitive boosting for classification of imbalanced data (2007, methodology): “For decision trees, one approach is to adjust the probabilistic estimate at the tree leaf [71, 105]; another approach is to develop new pruning techniques [105].”
Outlier Analysis (2013, methodology): “The work in [497] proposes methods to learn the costs directly in a data driven manner.”
C4.5 and Imbalanced Data sets: Investigating the eect of sampling method, probabilistic estimate, and decision tree structure (2003, methodology): “However, Laplace estimates might not be very appropriate for highly imbalanced data sets (Zadrozny & Elkan, 2001).”
Decision trees: a recent overview (2011, methodology): “This cost-sensitive learning method is called direct cost-sensitive learning ( Zadrozny and Elkan 2001 ).”
A study of the behavior of several methods for balancing machine learning training data (2004, result): “Although some research papers state that pruning might be helpful with imbalanced data sets in some circumstances [4], other papers indicate that when target misclassification costs or class distributions are unknown, then pruning should be avoided [26; 2].”
Handling imbalanced datasets: A review (2006, background): “Class imbalances are not the only problem to contend with: the distribution of the data within each class is also relevant (between-class versus within-class imbalance) [17], [34].”
An insight into classification with imbalanced data: Empirical results and current trends on using data intrinsic characteristics (2013, methodology): “Algorithmic modification: This procedure is oriented towards the adaptation of base learning methods to be more attuned to class imbalance issues [147] 3.”
When Training and Test Sets are Different: Characterising Learning Transfer (2013, background): “As costs and probabilities are intimately linked, the paper [32] discusses how to jointly deal with these unknowns.”
Adaptive Machine Learning for Credit Card Fraud Detection (2015, methodology): “Following the same idea Zadrozny and Elkan [142] use a classifier to predict P(s = 1|x) and then incorporate this probability estimate as input feature in a new classifier used to predict P(y|x), but there is no theoretical guarantee that this method should correct SSB.”