# AdaCost: Misclassification Cost-Sensitive Boosting

@inproceedings{Fan1999AdaCostMC, title={AdaCost: Misclassification Cost-Sensitive Boosting}, author={Wei Fan and S. Stolfo and Junxin Zhang and P. Chan}, booktitle={ICML}, year={1999} }

AdaCost, a variant of AdaBoost, is a misclassification cost-sensitive boosting method. It uses the cost of misclassifications to update the training distribution on successive boosting rounds. The purpose is to reduce the cumulative misclassification cost more than AdaBoost. We formally show that AdaCost reduces the upper bound of cumulative misclassification cost of the training set. Empirical evaluations have shown significant reduction in the cumulative misclassification cost over AdaBoost… Expand

#### Supplemental Presentations

#### 609 Citations

Distributed Cost Boosting and Bounds on Mis-classification Cost

- Computer Science
- 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'06)
- 2006

A novel algorithm, termed "distributed cost boosting" (DCB), that incorporates the cost metric into the error computation step, rather than merely adjusting the distribution update step is presented, which produces large empirical reductions in total cost when compared to current state-ofthe- art cost-sensitive Adaboost variants. Expand

Cost-sensitive Boosting with p-norm Loss Functionsand its Applications

- Computer Science
- 2008

This paper introduces a family of novel boosting methods for cost-sensitive classification by applying the theory of gradient boosting to p-norm based cost functionals, and establishes theoretical guarantees as well as their empirical advantage over existing algorithms. Expand

An Empirical Evaluation of Adaboost Extensions for Cost-Sensitive Classification

- Computer Science
- 2012

All the cost-sensitive boosters are studied and five new extensions are proposed and their results are compared in this paper and a few future extensions are notified. Expand

Cost-sensitive boosting algorithms as gradient descent

- Computer Science
- 2008 IEEE International Conference on Acoustics, Speech and Signal Processing
- 2008

It is shown that several typical CSB methods can also be view as gradient descent for minimizing a unified objective function and deduce a general greedy boosting procedure. Expand

Fisher Consistent Cost Sensitive Boosting Algorithm

- Computer Science
- 2013

This paper studies the problem of designing cost sensitive Boosting algorithms strictly under Boosting theory, and two new loss functions are constructed by making exponential loss and logit loss cost sensitive. Expand

An Empirical evaluation of CostBoost Extensions for Cost-Sensitive Classification

- Computer Science
- Compute '15
- 2015

Boosting is used for combining multiple classifiers using bagging, boosting or stacking to predict group membership for data samples and three new extensions of CostBoost algorithm CBE1, CBE2 and CBE3 are proposed and compared with existing cost based boosting classifiers. Expand

Parameter Inference of Cost-Sensitive Boosting Algorithms

- Computer Science
- MLDM
- 2005

Three versions of cost-sensitive AdaBoost algorithms where the parameters for sample weight updating are induced are proposed and their identification abilities on the small classes are tested on four “real world” medical data sets taken from UCI Machine Learning Database based on F-measure. Expand

Cost-Sensitive Boosting: Fitting an Additive Asymmetric Logistic Regression Model

- Computer Science
- ACML
- 2009

A framework from a statistical insight that can embody almost all existing cost-sensitive boosting algorithms: fitting an additive asymmetric logistic regression model by stage-wise optimization of certain criterions is developed. Expand

Precision-based Boosting

- Computer Science
- AAAI
- 2021

This paper proposes a generic refinement of all of these AdaBoost variants by using classspecific error rates, which is guaranteed to boost weak learners into strong learners. Expand

Cost-sensitive Boosting Cost-sensitive Boosting

- 2007

A novel framework, based on the statistical interpretation of boosting, is proposed for the design of cost sensitive boosting algorithms. It is argued that, although predic-tors produced with… Expand

#### References

SHOWING 1-10 OF 19 REFERENCES

Boosting Trees for Cost-Sensitive Classifications

- Computer Science
- ECML
- 1998

Two boosting techniques for cost-sensitive tree classifications in the situation where misclassification costs change very often are explored, demonstrating robustness of the induced model against cost changes. Expand

Boosting and Rocchio applied to text filtering

- Computer Science
- SIGIR '98
- 1998

We discuss two learning algorithms for text filtering: modified Rocchio and a boosting algorithm called AdaBoost. We show how both algorithms can be adapted to maximize any general utility matrix… Expand

Optimizing Classifers for Imbalanced Training Sets

- Computer Science
- NIPS
- 1998

This paper investigates the implications of results for the case of imbalanced datasets and develops two approaches to setting the threshold, incorporated into ThetaBoost, a boosting algorithm for dealing with unequal loss functions. Expand

Improved Boosting Algorithms Using Confidence-rated Predictions

- Computer Science
- COLT' 98
- 1998

We describe several improvements to Freund and Schapire's AdaBoost boosting algorithm, particularly in a setting in which hypotheses may assign confidences to each of their predictions. We give a… Expand

Cost-Sensitive Classification: Empirical Evaluation of a Hybrid Genetic Decision Tree Induction Algorithm

- Computer Science
- J. Artif. Intell. Res.
- 1995

This paper introduces ICET, a new algorithm for cost-sensitive classification that uses a genetic algorithm to evolve a population of biases for a decision tree induction algorithm and establishes that ICET performs significantly better than its competitors. Expand

Credit Card Fraud Detection Using Meta-Learning: Issues and Initial Results 1

- Computer Science
- 1997

It is argued that, for the fraud detection domain, fraud catching rate and false alarm rate are better metrics than the overall accuracy when evaluating the learned fraud classifiers, and that given a skewed distribution in the original data, artificially more balanced training data leads to better classifiers. Expand

Toward Scalable Learning with Non-Uniform Class and Cost Distributions: A Case Study in Credit Card Fraud Detection

- Computer Science
- KDD
- 1998

A multi-classifier meta-learning approach to address very large databases with skewed class distributions and non-uniform cost per error and empirical results indicate that the approach can significantly reduce loss due to illegitimate transactions. Expand

Analysis and Visualization of Classifier Performance: Comparison under Imprecise Class and Cost Distributions

- Computer Science
- KDD
- 1997

The ROC convex hull method combines techniques from ROC analysis, decision analysis and computational geometry, and adapts them to the particulars of analyzing learned classifiers to present a method for the comparison of classifier performance that is robust to imprecise class distributions and misclassification costs. Expand

A decision-theoretic generalization of on-line learning and an application to boosting

- Computer Science
- EuroCOLT
- 1995

The model studied can be interpreted as a broad, abstract extension of the well-studied on-line prediction model to a general decision-theoretic setting, and the multiplicative weightupdate Littlestone Warmuth rule can be adapted to this model, yielding bounds that are slightly weaker in some cases, but applicable to a considerably more general class of learning problems. Expand

A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting

- Computer Science, Mathematics
- COLT 1997
- 1997

The model studied can be interpreted as a broad, abstract extension of the well-studied on-line prediction model to a general decision-theoretic setting, and it is shown that the multiplicative weight-update Littlestone?Warmuth rule can be adapted to this model, yielding bounds that are slightly weaker in some cases, but applicable to a considerably more general class of learning problems. Expand