Detecting Anomalies in Sequential Data Augmented with New Features

Xiangzeng Kong, Y Bi, David H. Glass

Research output: Contribution to journal › Article

Abstract

This paper presents a new weighted local outlier factor method for anomaly detection, which is underpinned with three novel components: (1) a piecewise linear representa-tion defined on the basis of the important points that consist of extreme points and additional points; (2) a set of new features which are used to identify anomalies given the new piecewise linear representation; (3) a weighting schema, assigning different weights to different features by accounting for the discriminant power of the features. The underlying idea of the proposed method is to characterize a time series with a set of four features and then discover abnormal changes by taking account of the close-ness of any data points augmented with the new features. The comparative experi-ments demonstrate that the proposed piecewise representation method has performed well in sequential time series data, and the weighted local outlier factor method has achieved better accuracy and RankPower in detecting anomalies from the same data sets in comparison with the conventional local outlier factor, normalized local outlier factor and HOT symbolic aggregate approximation methods.
LanguageEnglish
JournalArtificial Intelligence Review
Publication statusAccepted/In press - 18 Nov 2018

Fingerprint

local factors
Time series
time series
weighting
Anomaly
Outliers

Keywords

  • Anomaly detection
  • sequential data
  • feature extraction
  • weighted local outlier factor

Cite this

Kong, Xiangzeng ; Bi, Y ; Glass, David H. / Detecting Anomalies in Sequential Data Augmented with New Features. In: Artificial Intelligence Review. 2018.
@article{e2c84aed427143b4a64098540501294a,
title = "Detecting Anomalies in Sequential Data Augmented with New Features",
abstract = "This paper presents a new weighted local outlier factor method for anomaly detection, which is underpinned with three novel components: (1) a piecewise linear representa-tion defined on the basis of the important points that consist of extreme points and additional points; (2) a set of new features which are used to identify anomalies given the new piecewise linear representation; (3) a weighting schema, assigning different weights to different features by accounting for the discriminant power of the features. The underlying idea of the proposed method is to characterize a time series with a set of four features and then discover abnormal changes by taking account of the close-ness of any data points augmented with the new features. The comparative experi-ments demonstrate that the proposed piecewise representation method has performed well in sequential time series data, and the weighted local outlier factor method has achieved better accuracy and RankPower in detecting anomalies from the same data sets in comparison with the conventional local outlier factor, normalized local outlier factor and HOT symbolic aggregate approximation methods.",
keywords = "Anomaly detection, sequential data, feature extraction, weighted local outlier factor",
author = "Xiangzeng Kong and Y Bi and Glass, {David H.}",
year = "2018",
month = "11",
day = "18",
language = "English",
journal = "Artificial Intelligence Review",
issn = "0269-2821",

}

Kong, X, Bi, Y & Glass, DH 2018, 'Detecting Anomalies in Sequential Data Augmented with New Features', Artificial Intelligence Review.

Detecting Anomalies in Sequential Data Augmented with New Features. / Kong, Xiangzeng; Bi, Y; Glass, David H.

In: Artificial Intelligence Review, 18.11.2018.

Research output: Contribution to journal › Article

TY - JOUR

T1 - Detecting Anomalies in Sequential Data Augmented with New Features

AU - Kong, Xiangzeng

AU - Bi, Y

AU - Glass, David H.

PY - 2018/11/18

Y1 - 2018/11/18

N2 - This paper presents a new weighted local outlier factor method for anomaly detection, which is underpinned with three novel components: (1) a piecewise linear representa-tion defined on the basis of the important points that consist of extreme points and additional points; (2) a set of new features which are used to identify anomalies given the new piecewise linear representation; (3) a weighting schema, assigning different weights to different features by accounting for the discriminant power of the features. The underlying idea of the proposed method is to characterize a time series with a set of four features and then discover abnormal changes by taking account of the close-ness of any data points augmented with the new features. The comparative experi-ments demonstrate that the proposed piecewise representation method has performed well in sequential time series data, and the weighted local outlier factor method has achieved better accuracy and RankPower in detecting anomalies from the same data sets in comparison with the conventional local outlier factor, normalized local outlier factor and HOT symbolic aggregate approximation methods.

AB - This paper presents a new weighted local outlier factor method for anomaly detection, which is underpinned with three novel components: (1) a piecewise linear representa-tion defined on the basis of the important points that consist of extreme points and additional points; (2) a set of new features which are used to identify anomalies given the new piecewise linear representation; (3) a weighting schema, assigning different weights to different features by accounting for the discriminant power of the features. The underlying idea of the proposed method is to characterize a time series with a set of four features and then discover abnormal changes by taking account of the close-ness of any data points augmented with the new features. The comparative experi-ments demonstrate that the proposed piecewise representation method has performed well in sequential time series data, and the weighted local outlier factor method has achieved better accuracy and RankPower in detecting anomalies from the same data sets in comparison with the conventional local outlier factor, normalized local outlier factor and HOT symbolic aggregate approximation methods.

KW - Anomaly detection

KW - sequential data

KW - feature extraction

KW - weighted local outlier factor

M3 - Article

JO - Artificial Intelligence Review

T2 - Artificial Intelligence Review

JF - Artificial Intelligence Review

SN - 0269-2821

ER -

Kong X, Bi Y, Glass DH. Detecting Anomalies in Sequential Data Augmented with New Features. Artificial Intelligence Review. 2018 Nov 18.