Mining concept-drifting data streams using ensemble classifiers

By Haixun Wang, Wei Fan, Philip S. Yu, Jiawei Han. 2003

1,530 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Networks and Communications

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 Computer Networks and Communications

Citations in the corpus

Ups and Downs: Modeling the Visual Evolution of Fashion Trends with One-Class Collaborative Filtering (2016, methodology): “Such learning algorithms include decision trees [37], SVMs [18], instance-based learning [1], etc.”
Measuring Catastrophic Forgetting in Neural Networks (2017, methodology): “Two methods that try to alleviate the memory usage problem are Accuracy Weighted Ensem-bles and Life-long Machine Learning (Wang et al. 2003; Ren et al. 2017).”
Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis (2019, methodology): “Another widely used method is ensembling multiple clas-siﬁers [5, 6, 48].”
RODEO: Replay for Online Object Detection (2020, neutral): “work sparsely updates weights to mitigate interference [8], 3) ensembling multiple classifiers together [9, 11, 43, 47, 53], and 4) rehearsal/replay models that store a subset of previous training inputs (or generate previous inputs) to mix with new examples when updating the network [4, 16, 17, 21,”
Adaptive Machine Learning for Credit Card Fraud Detection (2015, methodology): “Alternatively a weighted ensemble approach can be used to include new rules while maintaining old rules [189].”
Learning under Concept Drift: an Overview (2010, background): “Often these blocks are non overlapping [142, 150, 148, 7, 115, 72, 81, 79, 95], sometimes overlapping [44].”
Mining data streams: a review (2005, background): “There is a real need inspired by the potential applications in astronomy and scientific laboratories [ 23 ] as well as business applications to address the above research problems.”
Classifier Ensembles for Changing Environments (2004, methodology): “[22] propose to evaluate all classifiers using the most recent “chunk” of data as the testing set.”
Life-long learning based on dynamic combination model (2017, methodology): “In AdaBoost/TradaBoost, Learn++ and AWE, models are combined linearly by setting weights to the them based on their performance.”
A survey on data stream clustering and classification (2015, methodology): “[98] proposed an accuracy-weighted ensemble (AWE) classifier for mining concept-drifting in data streams.”
Ensemble learning for data stream analysis: A survey (2017, methodology): “The key idea of AWE is to assign weights to each classiﬁer of the ensemble based on their prediction error on the newest training chunk.”
A Comprehensive Survey of Data Mining-based Fraud Detection Research (2010, background): “…survey papers and bibliographies include outlier detection (Hodge and Austin, 2004), skewed/imbalanced/rare classes 4 (Weiss, 2004), sampling (Domingos et al , 2002), cost sensitive learning 5 , stream mining 6 , graph mining (Washio and Motoda, 2003), and scalability (Provost and Kolluri, 1999).”
Concept drift detection via competence models (2014, methodology): “These include rule-based learning [4, 8, 12], decision trees and their incremental versions [5, 13, 14], info-fuzzy networks [15], clustering [16], support vector machines [17], and case-based reasoning (CBR) [11, 18-20].”
Learning With Feature Evolvable Streams (2017, methodology): “For the latter, various ensemble methods have been proposed including Online Bagging & Boosting [23], Weighted Ensemble Classifiers [31, 21], Adapted One-vs-All Decision Trees (OVA) [12] and Meta-knowledge Ensemble [34].”
Regularizing Deep Convolutional Neural Networks with a Structured Decorrelation Constraint (2016, methodology): “To reduce overfitting, several regularizers, optimization methods and network structures have been proposed, including Early Stopping [13], Weight Decay [14], Model Ensemble [15–17], Data Augmentation [18], Dropout [19, 20], DropConnect [21], Maxout [22], Batch Normalization [23], Residual Networks…”
Classifier Ensembles for Detecting Concept Change in Streaming Data: Overview and Perspectives (2008, methodology): “Included in this group are most ensemble methods for changing environments: Winnow variants [24, 28], AdaBoost variants [5, 30], “replace-the-loser” approaches [21,22,34,35,39].”