Dynamic Programming for Bayesian Logistic Regression Learning under Concept Drift

Turkov, Pavel A.; Krasotkina, Olga; Mottl, Vadim

doi:10.1007/978-3-642-45062-4_26

Cited by 8 publications

(2 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, the newly proposed single classifier and active detection algorithms generally contain a novel detection mechanism for environment changes, such as CUSUM algorithm [10,11] based on control charts and OLIN algorithm [12] based on the error confidence interval and other methods based on statistical techniques [13]. These algorithms regard the classification learning problem in the non-stationary environment as a kind of prediction problem of environment changes [14] or as the minimization problem of the errors penalty function of classifier [15,16].…”

Section: Relevant Algorithmsmentioning

confidence: 99%

A fast learn++.NSE classification algorithm based on weighted moving average

Shen¹,

Zhu²,

Du³

et al. 2018

Filomat

View full text Add to dashboard Cite

Current researches of incremental classification learning algorithms mainly focus on learning from data in a stationary environment. The incremental learning in a non-stationary environment (NSE), where the underlying data probability distribution changes over time, however, has received much less attentions despite the abundant real applications have generated the long-term and cumulative big data in NSE. Thus, the incremental learning in NSE has gradually received extensive attentions. Nevertheless, the popular incremental classification learning algorithms currently for NSE such as SEA and DWM generally place strict restrictions on the changes. These algorithms can only deal with gradual drift and noncyclical and no new category situations. Therefore, it is highly necessary to develop a novel efficient incremental classification learning algorithm for the gradually cumulative big data in complex NSE. The recently proposed Learn++.NSE algorithm is an important research achievement in this field. However, the vote weight of each base-classifier of the Learn++.NSE depends on its whole error rates in the environments experienced. Therefore, the classification learning efficiency of the Learn++.NSE should be further improved. A novel fast Learn++.NSE algorithm based on weighted moving average (WMA-Learn++.NSE) is presented in this paper, which computes the weighted average of error rates using the sliding window technology to optimize the weight calculation. By only using the recent classification error rates of each base-classifier inside the sliding window to calculate the vote weight, the WMA-Learn++.NSE accelerates the compute of vote weight and improves the efficiency of classification learning. The verification experiments and performance analyses on both synthetic and real data set are presented in this paper. The experimental results show that the WMA-Learn++.NSE can achieve a higher execution efficiency compared to the Learn++.NSE in getting the equivalent classification correct rate.

show abstract

Section: Relevant Algorithmsmentioning

confidence: 99%

A fast learn++.NSE classification algorithm based on weighted moving average

Shen¹,

Zhu²,

Du³

et al. 2018

Filomat

View full text Add to dashboard Cite

show abstract

“…For a more detailed overview of the results in this area, including online incremental ensembles, readers may refer to monograph [7] and review [8]. Bayesian logistic regression was used in [9] to handle drifting concepts in terms of dynamical programming. Concept drifting handling is close to methodology of on-line random forest algorithm [10] where ideas from on-line bagging, extremely randomised forests and on-line decision tree growing procedure are employed.…”

Section: Related Workmentioning

confidence: 99%

Random Forest Based Approach for Concept Drift Handling

Жуков

Sidorov

Foley

2017

Communications in Computer and Information Science

View full text Add to dashboard Cite

Abstract. Concept drift has potential in smart grid analysis because the socio-economic behaviour of consumers is not governed by the laws of physics. Likewise there are also applications in wind power forecasting. In this paper we present decision tree ensemble classification method based on the Random Forest algorithm for concept drift. The weighted majority voting ensemble aggregation rule is employed based on the ideas of Accuracy Weighted Ensemble (AWE) method. Base learner weight in our case is computed for each sample evaluation using base learners accuracy and intrinsic proximity measure of Random Forest. Our algorithm exploits both temporal weighting of samples and ensemble pruning as a forgetting strategy. We present results of empirical comparison of our method with оriginal random forest with incorporated "replace-thelooser" forgetting andother state-of-the-art concept-drfit classifiers like AWE2.

show abstract

A Hybrid Distance-Based and Naive Bayes Online Classifier

Jȩdrzejowicz

Jędrzejowicz

2015

Computational Collective Intelligence

View full text Add to dashboard Cite

Dynamic Programming for Bayesian Logistic Regression Learning under Concept Drift

Cited by 8 publications

References 13 publications

A fast learn++.NSE classification algorithm based on weighted moving average

A fast learn++.NSE classification algorithm based on weighted moving average

Random Forest Based Approach for Concept Drift Handling

A Hybrid Distance-Based and Naive Bayes Online Classifier

Contact Info

Product

Resources

About