|
Title:
|
A PIPELINE-BASED DATA STREAM MINING SYSTEM FOR OUTLIER DETECTION IN EVOLVING STREAMS |
|
Author(s):
|
Maxim Shcherbakov, Yury Timofeev, Alexander Saprykin, Vyacheslav Trushin, Anton Tyukov, Valeriy Kamaev |
|
ISBN:
|
978-972-8939-93-9 |
|
Editors:
|
António Palma dos Reis and Ajith P. Abraham |
|
Year:
|
2013 |
|
Edition:
|
Single |
|
Keywords:
|
Data stream mining, pipeline-based design, outlier detection |
|
Type:
|
Short Paper |
|
First Page:
|
84 |
|
Last Page:
|
88 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
The issue of outlier detection in evolving data stream can be indicated as a crucial for control and decision making. Evolving data streams are obtained via observation of the systems with different operating mode. Hence the normal value in the data stream fetched during one operation mode might be outliers in the other mode. The contribution of the paper is the design of the new pipeline-based data stream mining system (Smarterdam) for outlier detection in evolving data streams. This system has the following features: (i) an adaptive on-line and off-line pipelines for outlier detection in evolving data streams and (ii) high-level NoSQL query language with JSON-like sets of parameters for outlier detection task in data streams. This concept allows to (i) run the outlier procedure without human intervention in the detection procedures and (ii) use different outlier detection techniques without affecting the performance of data stream processing. |
|
|
|
|
|
|
