RDDM: Reactive drift detection method

被引:125
作者
Barros, Roberto S. M. [1 ]
Cabral, Danilo R. L. [1 ]
Goncalves, Paulo M., Jr. [2 ]
Santos, Silas G. T. C. [1 ]
机构
[1] Univ Fed Pernambuco, Ctr Informat, Cidade Univ, BR-50740560 Recife, PE, Brazil
[2] Inst Fed Educ Ciencia & Tecnol Pernambuco, Cidade Univ, BR-50740540 Recife, PE, Brazil
关键词
Concept drift; Drift detection methods; Data stream; Online learning; ONLINE;
D O I
10.1016/j.eswa.2017.08.023
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Concept drift detectors are online learning software that mostly attempt to estimate the drift positions in data streams in order to modify the base classifier after these changes and improve accuracy. This is very important in applications such as the detection of anomalies in TCP/IP traffic and/or frauds in financial transactions. Drift Detection Method (DDM) is a simple, efficient, well-known method whose performance is often impaired when the concepts are very long. This article proposes the Reactive Drift Detection Method (RDDM), which is based on DDM and, among other modifications, discards older instances of very long concepts aiming to detect drifts earlier, improving the final accuracy. Experiments run in MOA, using abrupt and gradual concept drift versions of different dataset generators and sizes (48 artificial datasets in total), as well as three real-world datasets, suggest RDDM beats the accuracy results of DDM, ECDD, and STEPD in most scenarios. (C) 2017 Elsevier Ltd. All rights reserved.
引用
收藏
页码:344 / 355
页数:12
相关论文
共 35 条
[11]  
Cormen Thomas H, 2009, Introduction to Algorithms
[12]   STATISTICAL-THEORY - THE PREQUENTIAL APPROACH [J].
DAWID, AP .
JOURNAL OF THE ROYAL STATISTICAL SOCIETY SERIES A-STATISTICS IN SOCIETY, 1984, 147 :278-292
[13]  
Demsar J, 2006, J MACH LEARN RES, V7, P1
[14]   A Selective Detector Ensemble for Concept Drift Detection [J].
Du, Lei ;
Song, Qinbao ;
Zhu, Lei ;
Zhu, Xiaoyan .
COMPUTER JOURNAL, 2015, 58 (03) :457-471
[15]   Online and Non-Parametric Drift Detection Methods Based on Hoeffding's Bounds [J].
Frias-Blanco, Isvani ;
del Campo-Avila, Jose ;
Ramos-Jimenez, Gonzalo ;
Morales-Bueno, Rafael ;
Ortiz-Diaz, Agustin ;
Caballero-Mota, Yaile .
IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING, 2015, 27 (03) :810-823
[16]  
Gama J, 2004, LECT NOTES ARTIF INT, V3171, P286
[17]   A Survey on Concept Drift Adaptation [J].
Gama, Joao ;
Zliobaite, Indre ;
Bifet, Albert ;
Pechenizkiy, Mykola ;
Bouchachia, Abdelhamid .
ACM COMPUTING SURVEYS, 2014, 46 (04)
[18]  
Garrido Teixeira de Carvalho Santos Silas, 2014, Machine Learning and Knowledge Discovery in Databases. European Conference, ECML PKDD 2014. Proceedings: LNCS 8726, P179, DOI 10.1007/978-3-662-44845-8_12
[19]   A comparative study on concept drift detectors [J].
Goncalves, Paulo M., Jr. ;
de Carvalho Santos, Silas G. T. ;
Barros, Roberto S. M. ;
Vieira, Davi C. L. .
EXPERT SYSTEMS WITH APPLICATIONS, 2014, 41 (18) :8144-8156
[20]   RCD: A recurring concept drift framework [J].
Goncalves, Paulo Mauricio, Jr. ;
Major de Barros, Roberto Souto .
PATTERN RECOGNITION LETTERS, 2013, 34 (09) :1018-1025