DDD: A New Ensemble Approach for Dealing with Concept Drift

Minku, Leandro L.; Yao, Xin

doi:10.1109/tkde.2011.58

Cited by 381 publications

(283 citation statements)

References 32 publications

Supporting

Mentioning

281

Contrasting

Unclassified

Order By: Relevance

“…This is not unexpected, because a learner cannot learn what has not been taught. Such a situation might be better handled by techniques for detecting concept drifts (Minku and Yao 2012b). Similarly, on time steps 30-37, both CC-RT0 and WC-RT win frequently, causing their weights to increase competitively during this period.…”

Section: Discussionmentioning

confidence: 99%

“…5 was performed, as both DCL and the RTs used in this study are deterministic and the datasets must have their order of examples fixed to represent the real online learning scenario of a company. This is a standard procedure when evaluating online learning approaches (Minku and Yao 2012b;Kolter and Maloof 2007).…”

Section: Methodsmentioning

confidence: 99%

“…It is included in the analysis to check whether DCL would be able to improve predictive performance over an existing dynamic adaptive approach and was first investigated in the SEE context in Minku and Yao (2012a). Different from DDD (Minku and Yao 2012b), DWM keeps base learners likely to represent several different concepts. This is more likely to be beneficial for SEE, where each concept is available for a short period of time and may reoccur in the future.…”

Section: Methodsmentioning

confidence: 99%

“…However, AddExp is less robust to noise. Another approach is Diversity for Dealing with Drifts (DDD) (Minku and Yao 2012b). Its main idea is that very highly diverse ensembles (whose base learners produce very different predictions from each other) are likely to present poor predictive performance under stable conditions, but may become useful when there are changes.…”

Section: Literature On Predictive Models For Non-stationary Enviromentioning

confidence: 99%

See 3 more Smart Citations

Which models of the past are relevant to the present? A software effort estimation approach to exploiting useful past models

Minku

Yao

2016

Autom Softw Eng

Self Cite

View full text Add to dashboard Cite

Software Effort Estimation (SEE) models can be used for decision-support by software managers to determine the effort required to develop a software project. They are created based on data describing projects completed in the past. Such data could include past projects from within the company that we are interested in (WC projects) and/or from other companies (cross-company, i.e., CC projects). In particular, the use of CC data has been investigated in an attempt to overcome limitations caused by the typically small size of WC datasets. However, software companies operate in non-stationary environments, where changes may affect the typical effort required to develop software projects. Our previous work showed that both WC and CC models of the past can become more or less useful over time, i.e., they can sometimes be helpful and sometimes misleading. So, how can we know if and when a model created based on past data represents well the current projects being estimated? We propose an approach called Dynamic Cross-company Learning (DCL) to dynamically identify which WC or CC past models are most useful for making predictions to a given company at the present. DCL automatically emphasizes the predictions given by these models in order to improve predictive performance. Our experiments comparing DCL against existing WC and CC approaches show that DCL is successful in improving SEE by A preliminary version of this work appeared as Minku and Yao (2012a

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

Section: Literature On Predictive Models For Non-stationary Enviromentioning

confidence: 99%

See 2 more Smart Citations

Which models of the past are relevant to the present? A software effort estimation approach to exploiting useful past models

Minku

Yao

2016

Autom Softw Eng

Self Cite

View full text Add to dashboard Cite

show abstract

“…It can offer partly help in detecting sudden changes to use small size chunks, while this may also damage the stability and computational costs performance of the ensemble. Simple incremental learning [9][10][11][12][13][14] keeps sensitivity to sudden changes to make self-adaption timely but is not enough for coping with gradual drifts as forgetting history data and sharp adaptation to newest status. In order to adapt to both sudden and gradual changes, it could be suitable to combine significant features from block-based ensembles and incremental learning approaches.…”

Section: Introductionmentioning

confidence: 99%

Online Active Learning with Drifted Data Streams Using Paired Ensemble Framework

et al. 2017

View full text Add to dashboard Cite

Abstract. In learning to classify data streams, it is impractical and expensive to label all of the instances. Online active learning over streaming data poses additional challenges for its increasing volumes and concept drifts. We propose a new online paired ensemble active learning framework consisting of a stable classifier and a timely substituted dynamic classifier to react to different types of concept drifts. Classifiers are built in block based way and will learn new instances incrementally online. According to a combination strategy of uncertainty strategy and random strategy, the decision whether to label the incoming instance for the updating of the stable classifier and the dynamic classifier will be made. Experimental evaluation results on real datasets show the advantage of the proposed work in comparison with other approaches.

show abstract

Distributed Online Learning and Stream Processing for a Smarter Planet

Turaga¹,

Schaar

2017

Fog for 5G and IoT

View full text Add to dashboard Cite

Smarter Planet applications for transportation, healthcare, energy and utilities have been enabled by the confluence of technological advancements in sensing, communication, and compute systems, and analysis algorithms. These applications have have several novel characteristics in terms of their distributed nature, their analysis, security and privacy needs, their performance scaling needs, and their dynamics -that have not been sufficiently addressed by prior centrallystore-process-and-analyze systems and algorithms.In this chapter we describe the emerging field of Stream Processing systems and its impact on the realization of these applications -and discuss how this is enabling fog computing in real-world implementations. We introduce the paradigm, its core constructs and capabilities, and examples of academic, commercial and open-source systems available for use. We also highlight examples of large-scale distributed applications built on these systems, and in use in current Smarter Planet applications. We then discuss how these systems enable several new directions

show abstract

DDD: A New Ensemble Approach for Dealing with Concept Drift

Cited by 381 publications

References 32 publications

Which models of the past are relevant to the present? A software effort estimation approach to exploiting useful past models

Which models of the past are relevant to the present? A software effort estimation approach to exploiting useful past models

Online Active Learning with Drifted Data Streams Using Paired Ensemble Framework

Distributed Online Learning and Stream Processing for a Smarter Planet

Contact Info

Product

Resources

About