Fast adaptive stacking of ensembles

Blanco, Isvani Frías; Verdecia-Cabrera, Alberto; Díaz, Agustín Alejandro Ortiz; Carvalho, André Carlos Ponce de Leon Ferreira

doi:10.1145/2851613.2851655

Cited by 33 publications

(21 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…By reducing the ADOB's requirement condition of allowing base classifiers to vote and changing the concept drift detection method internally used, BOLE can cope with different types and amounts of concept drift more effectively. Frias-Blanco et al [19] proposed the Fast Adaptive Stacking of Ensembles (FASE), which constructs metainstance M = (ĉ 1 ,ĉ 2 , . .…”

Section: A Data Stream Classification Ensemble Learningmentioning

confidence: 99%

“…end if [11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26][27], and call the process DealInstance to learn the data of the circular array A in turn and cache the new instances (line [11][12][13][14]. Whenever array A is filled again by new instances, CreateNewBaseClassifier is called After the reading of the data stream is finished, the array still contains I instances that need to be predicted and learned, so DealInstance is called to process them (line 28-31).…”

Section: B Ensemble Process Of Online Active Learningmentioning

confidence: 99%

“…On one hand, it is compared with the four representative active learning strategies (ALStrategies) described in [29], including Random Sampling (Random), Variable Uncertainty Strategy (VarUn), Uncertainty Strategy with Randomization (RanVarUn), and Selective Sample (SelSample). On the other hand, it is compared with supervised ensemble learning methods, i.e., AUE [3], BOLE [18], and FASE [19], to observe the performance of the proposed method relative to the supervised learning. All experiments were conducted with the Massive Online Analysis (MOA) data stream software suite [37], an open source software environment for data stream mining, which includes a series of implemented data stream mining algorithms and evaluation methods.…”

Section: Experimental Evaluationmentioning

confidence: 99%

“…Generally, ensemble methods can be divided into online ensembles and block-based ensembles in accordance with the different processing method of the instances. For the online ensemble [14]- [19], newly arriving instances are processed one by one, and the base classifier 2162-237X © 2018 IEEE. Translations and content mining are permitted for academic research only.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Online Active Learning Ensemble Framework for Drifted Data Streams

Shan

Zhang

Liu

et al. 2019

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

In practical applications, data stream classification faces significant challenges, such as high cost of labeling instances and potential concept drifting. We present a new online active learning ensemble framework for drifting data streams based on a hybrid labeling strategy that includes the following: 1) an ensemble classifier, which consists of a long-term stable classifier and multiple dynamic classifiers (a multilevel sliding window model is used to create and update the dynamic classifiers to effectively process both the gradual drift type and sudden drift type data stream) and 2) active learning, which takes a nonfixed labeling budget, supports on-demand request labeling, and adopts an uncertainty strategy and random strategy to label instances. The decision threshold of the uncertainty strategy is adjusted dynamically, i.e., when concept drift occurs, the threshold is gradually reduced to query the most uncertain instances in priority to reduce the request expense as much as possible. Experiments on synthetic and real data sets show that precise prediction accuracy can be obtained by the proposed method without increasing the total cost of labeling, and that the labeling cost can be dynamically allocated according to the concept drift.

show abstract

Section: A Data Stream Classification Ensemble Learningmentioning

confidence: 99%

Section: B Ensemble Process Of Online Active Learningmentioning

confidence: 99%

Section: Experimental Evaluationmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Online Active Learning Ensemble Framework for Drifted Data Streams

Shan

Zhang

Liu

et al. 2019

IEEE Trans. Neural Netw. Learning Syst.

View full text Add to dashboard Cite

show abstract

“…Se utilizaron 10 clasificadores base en todos los algoritmos. Como clasificador base se eligió Naive Bayes por ser uno de los algoritmos más exitosos para aprender de los flujos de datos [22,7,14], además tiene un costo computacional bajo y una semántica clara.…”

Section: Configuración De Los Experimentosunclassified