|
Activity Number:
|
249
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, July 31, 2007 : 8:30 AM to 10:20 AM
|
|
Sponsor:
|
Section on Nonparametric Statistics
|
| Abstract - #309814 |
|
Title:
|
Statistical Approaches to Mining Multivariate Datastreams
|
|
Author(s):
|
Eric Vance*+
|
|
Companies:
|
Duke University
|
|
Address:
|
114 Old Chemistry Building, Durham, NC, 27708,
|
|
Keywords:
|
data streams ; depth-based partitioning
; multivariate histograms ; Mahalanobis distance
|
|
Abstract:
|
Datastreams are fast moving, rapidly accumulating sequences of data that are a predominant source of information today. Monitoring these datastreams for anomalous behavior faces special challenges---rapid rate of accumulation; massive size and complexity; and brief access to the raw data. An effective strategy to address these issues is to first partition the attribute space into meaningful classes, and then use the class-summaries to analyze the streams further. In the context of two real-life applications we discuss a variety of data partitioning schemes and compare their performance in answering a fundamental question in the analysis of datastreams: Has the statistical behavior of the datastream changed? If so, in what way?
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2007 program |
