Evaluating boosting algorithms to classify rare classes: comparison and improvements

Joshi, Manjunath V.; Kumar, Vipin; Agarwal, R.

doi:10.1109/icdm.2001.989527

Cited by 189 publications

(123 citation statements)

References 6 publications

Supporting

Mentioning

121

Contrasting

Unclassified

Order By: Relevance

“…proposed that the AdaCost algorithm is one of the representative methods of cost-sensitive learning. Joshi et al [7]. showed that the AdaCost algorithm has high accuracy and recall rate for a few class samples.…”

Section: Related Work and Problemsmentioning

confidence: 99%

A Comprehensive Integration Method Based on Unbalanced Data Classification Problems

Lian

Zhang

Lin

et al. 2018

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Abstract. This paper first introduces the characteristics of non-equilibrium data and the main problems brought by its classification. Based on this, it proposes a comprehensive integration method based on non-equilibrium data classification problems, and the method improvement background and strategy. The construction of base classifiers, and the selective integration methods of base classifiers are described in detail. Finally, the method proposed in this chapter is verified by experiments. The experimental results show that the proposed method is effective.

show abstract

Section: Related Work and Problemsmentioning

confidence: 99%

A Comprehensive Integration Method Based on Unbalanced Data Classification Problems

Lian

Zhang

Lin

et al. 2018

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

show abstract

“…Several boosting algorithms have been developed since the preliminary work by Schapire 31 , which include cost-sensitive versions 32,33 and those which can deliver confidence approximations in their forecasts 34 . The most widely used variant of boosting methods is AdaBoost 35 .…”

Section: Constructingmentioning

confidence: 99%

Genome-Scale Classification of Recombinant and Non-Recombinant HIV-1 Sequences Using Artificial Neural Network Ensembles

Dwivedi

Chouhan

2016

Current Science

View full text Add to dashboard Cite

Genetic recombination and high rate of mutations in the HIV-1 genome increase the diversity of HIV-1, which allows viruses to escape more easily from host immune system or develop resistance for antiretroviral drugs. Consequently, it is indispensable to devise an effective method for recognition of recombination in HIV-1 strains. This article presents ensemble models of artificial neural network for the classification of recombinant and non-recombinant sequences of HIV-1 genome. We have evaluated the performance of these ensemble models using different classification measurements like specificity, sensitivity and classification accuracy. Furthermore, model performance was measured on receiver operating curve and using calibration graph. High classification accuracy up to 93.43% was achieved on tenfold cross validation.

show abstract

“…Therefore, several common performance metrics were applied, such as Recall (1), Precision (2), FP-rate (3), and an F-score (4) [6], [7]. Also, a Receiver Operating Characteristics (ROC) analysis [8] was applied.…”

Section: Classifier Performancementioning

confidence: 99%

Using Real-Valued Meta Classifiers to Integrate and Contextualize Binding Site Predictions

Robinson

Sharabi

Sun

et al. 2007

Adaptive and Natural Computing Algorithms

View full text Add to dashboard Cite

Abstract. Currently the best algorithms for transcription factor binding site predictions are severely limited in accuracy. However, a non-linear combination of these algorithms could improve the quality of predictions. A support-vector machine was applied to combine the predictions of 12 key real valued algorithms. The data was divided into a training set and a test set, of which two were constructed: filtered and unfiltered. In addition, a different "window" of consecutive results was used in the input vector in order to contextualize the neighbouring results. Finally, classification results were improved with the aid of under and over sampling techniques. Our major finding is that we can reduce the False-Positive rate significantly. We also found that the bigger the window, the higher the F-score, but the more likely it is to make a false positive prediction, with the best trade-off being a window size of about 7.

show abstract

Evaluating boosting algorithms to classify rare classes: comparison and improvements

Cited by 189 publications

References 6 publications

A Comprehensive Integration Method Based on Unbalanced Data Classification Problems

A Comprehensive Integration Method Based on Unbalanced Data Classification Problems

Genome-Scale Classification of Recombinant and Non-Recombinant HIV-1 Sequences Using Artificial Neural Network Ensembles

Using Real-Valued Meta Classifiers to Integrate and Contextualize Binding Site Predictions

Contact Info

Product

Resources

About