Avoiding Anomalies in Data Stream Learning

D 2013

Avoiding Anomalies in Data Stream Learning

GAMA, Joao, Petr KOSINA and Ezilda ALMEIDA

Basic information

Original name

Avoiding Anomalies in Data Stream Learning

Authors

GAMA, Joao (620 Portugal), Petr KOSINA (203 Czech Republic, guarantor, belonging to the institution) and Ezilda ALMEIDA (620 Portugal)

Edition

Berlin Heidelberg, Discovery Science, Proceedings of 16th International Conference DS 2013, p. 49-63, 15 pp. 2013

Publisher

Springer

Other information

Language

English

Type of outcome

Stať ve sborníku

Field of Study

10201 Computer sciences, information science, bioinformatics

Country of publisher

Germany

Confidentiality degree

není předmětem státního či obchodního tajemství

Publication form

printed version "print"

References:

URL

Impact factor

Impact factor: 0.402 in 2005

RIV identification code

RIV/00216224:14330/13:00070032

Organization unit

Faculty of Informatics

ISBN

978-3-642-40896-0

ISSN

DOI

http://dx.doi.org/10.1007/978-3-642-40897-7_4

UT WoS

000340562100004

Keywords in English

Data Streams; Rule Learning; Anomaly Detection

Abstract

V originále

The presence of anomalies in data compromises data quality and can reduce the effectiveness of learning algorithms. Standard data mining methodologies refer to data cleaning as a pre-processing before the learning task. The problem of data cleaning is exacerbated when learning in the computational model of data streams. In this paper we present a streaming algorithm for learning classification rules able to detect contextual anomalies in the data. Contextual anomalies are surprising attribute values in the context defined by the conditional part of the rule. For each example we compute the degree of anomaliness based on the probability of the attribute-values given the conditional part of the rule covering the example. The examples with high degree of anomaliness are signaled to the user and not used to train the classifier. The experimental evaluation in real-world data sets shows the ability to discover anomalous examples in the data. The main advantage of the proposed method is the ability to inform the context and explain why the anomaly occurs.

Links

LG13010, research and development project

Name: Zastoupení ČR v European Research Consortium for Informatics and Mathematics (Acronym: ERCIM-CZ)

Investor: Ministry of Education, Youth and Sports of the CR

MUNI/A/0758/2011, interní kód MU

Name: Zapojení studentů Fakulty informatiky do mezinárodní vědecké komunity (Acronym: SKOMU)

Investor: Masaryk University, Category A

Citovat

GAMA, Joao, Petr KOSINA and Ezilda ALMEIDA. Avoiding Anomalies in Data Stream Learning. In Johannes Furnkranz, Eyke Hullermeier,Tomoyuki Higuchi. Discovery Science, Proceedings of 16th International Conference DS 2013. Berlin Heidelberg: Springer, 2013, p. 49-63. ISBN 978-3-642-40896-0. Available from: https://dx.doi.org/10.1007/978-3-642-40897-7_4.

@inproceedings{1128372,
   author = {Gama, Joao and Kosina, Petr and Almeida, Ezilda},
   address = {Berlin Heidelberg},
   booktitle = {Discovery Science, Proceedings of 16th International Conference DS 2013},
   doi = {http://dx.doi.org/10.1007/978-3-642-40897-7_4},
   editor = {Johannes Furnkranz, Eyke Hullermeier,Tomoyuki Higuchi},
   keywords = {Data Streams; Rule Learning; Anomaly Detection},
   howpublished = {tištěná verze "print"},
   language = {eng},
   location = {Berlin Heidelberg},
   isbn = {978-3-642-40896-0},
   pages = {49-63},
   publisher = {Springer},
   title = {Avoiding Anomalies in Data Stream Learning},
   url = {http://link.springer.com/chapter/10.1007%2F978-3-642-40897-7_4},
   year = {2013}
}

TY  - JOUR
ID  - 1128372
AU  - Gama, Joao - Kosina, Petr - Almeida, Ezilda
PY  - 2013
TI  - Avoiding Anomalies in Data Stream Learning
PB  - Springer
CY  - Berlin Heidelberg
SN  - 9783642408960
KW  - Data Streams
KW  - Rule Learning
KW  - Anomaly Detection
UR  - http://link.springer.com/chapter/10.1007%2F978-3-642-40897-7_4
L2  - http://link.springer.com/chapter/10.1007%2F978-3-642-40897-7_4
N2  - The presence of anomalies in data compromises data quality and can reduce the effectiveness of learning algorithms. Standard data mining methodologies refer to data cleaning as a pre-processing before the learning task. The problem of data cleaning is exacerbated when learning in the computational model of data streams. In this paper we present a streaming algorithm for learning classification rules able to detect contextual anomalies in the data. Contextual anomalies are surprising attribute values in the context defined by the conditional part of the rule. For each example we compute the degree of anomaliness based on the probability of the attribute-values given the conditional part of the rule covering the example. The examples with high degree of anomaliness are signaled to the user and not used to train the classifier. The experimental evaluation in real-world data sets shows the ability to discover anomalous examples in the data. The main advantage of the proposed method is the ability to inform the context and explain why the anomaly occurs.
ER  -

GAMA, Joao, Petr KOSINA and Ezilda ALMEIDA. Avoiding Anomalies in Data Stream Learning. In Johannes Furnkranz, Eyke Hullermeier,Tomoyuki Higuchi. \textit{Discovery Science, Proceedings of 16th International Conference DS 2013}. Berlin Heidelberg: Springer, 2013, p.~49-63. ISBN~978-3-642-40896-0. Available from: https://dx.doi.org/10.1007/978-3-642-40897-7\_{}4.

Detailed Information on Publication Record