A Time-weighted average-based PAA representation for time series symbolization

Yahyia Benyahmed, Azuraliza Abu Bakar, Abdul Razak Hamdan, Sharifah Mastura Syed Abdullah

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

The operation of time series analysis to effectively manage the large amounts of data with high dimensional became an important research problem. Choose effective and scalable algorithms for appropriate representation of data is another challenge. A lot of high-level representations of the time series have been proposed for data extraction, such as spectral transfers, wavelets, piecewise polynomial, symbolic models, etc. One of the methods is Piecewise Aggregate Approximation (PAA) which minimizes dimensionality by the mean values of equal-sized frames, but this focus on mean value takes into consideration only the central tendency and not the dispersion present in each segment, which may lead to some important patterns being missed in some time series data sets. We propose method based on Time-Weighted Average for Symbolic Aggregate approximation method (TWA_SAX) compare its performance with some current methods. TWA_SAX is enables raw data to be specifically compared to the minimized representation and, at the same time, ensures reduced limits to Euclidean distance. It can be utilized to generate quicker, more precise algorithms for similarity searches, which improves the preciseness of time series representation through enabling better tightness of the lower bound.

Original languageEnglish
Pages (from-to)1-15
Number of pages15
JournalInternational Journal of Advances in Soft Computing and its Applications
Volume7
Issue number3
Publication statusPublished - 2015

Fingerprint

Time series
Time series analysis
Polynomials

Keywords

  • Dimensionality reduction
  • Discretize
  • Piecewise aggregate approximation
  • Symbolic
  • Time series

ASJC Scopus subject areas

  • Computer Science Applications

Cite this

@article{dacc6a37099d42958f7bdb2d28bb5533,
title = "A Time-weighted average-based PAA representation for time series symbolization",
abstract = "The operation of time series analysis to effectively manage the large amounts of data with high dimensional became an important research problem. Choose effective and scalable algorithms for appropriate representation of data is another challenge. A lot of high-level representations of the time series have been proposed for data extraction, such as spectral transfers, wavelets, piecewise polynomial, symbolic models, etc. One of the methods is Piecewise Aggregate Approximation (PAA) which minimizes dimensionality by the mean values of equal-sized frames, but this focus on mean value takes into consideration only the central tendency and not the dispersion present in each segment, which may lead to some important patterns being missed in some time series data sets. We propose method based on Time-Weighted Average for Symbolic Aggregate approximation method (TWA_SAX) compare its performance with some current methods. TWA_SAX is enables raw data to be specifically compared to the minimized representation and, at the same time, ensures reduced limits to Euclidean distance. It can be utilized to generate quicker, more precise algorithms for similarity searches, which improves the preciseness of time series representation through enabling better tightness of the lower bound.",
keywords = "Dimensionality reduction, Discretize, Piecewise aggregate approximation, Symbolic, Time series",
author = "Yahyia Benyahmed and {Abu Bakar}, Azuraliza and Hamdan, {Abdul Razak} and {Syed Abdullah}, {Sharifah Mastura}",
year = "2015",
language = "English",
volume = "7",
pages = "1--15",
journal = "International Journal of Advances in Soft Computing and its Applications",
issn = "2074-8523",
publisher = "International Center for Scientific Research and Studies (ICSRS)",
number = "3",

}

TY - JOUR

T1 - A Time-weighted average-based PAA representation for time series symbolization

AU - Benyahmed, Yahyia

AU - Abu Bakar, Azuraliza

AU - Hamdan, Abdul Razak

AU - Syed Abdullah, Sharifah Mastura

PY - 2015

Y1 - 2015

N2 - The operation of time series analysis to effectively manage the large amounts of data with high dimensional became an important research problem. Choose effective and scalable algorithms for appropriate representation of data is another challenge. A lot of high-level representations of the time series have been proposed for data extraction, such as spectral transfers, wavelets, piecewise polynomial, symbolic models, etc. One of the methods is Piecewise Aggregate Approximation (PAA) which minimizes dimensionality by the mean values of equal-sized frames, but this focus on mean value takes into consideration only the central tendency and not the dispersion present in each segment, which may lead to some important patterns being missed in some time series data sets. We propose method based on Time-Weighted Average for Symbolic Aggregate approximation method (TWA_SAX) compare its performance with some current methods. TWA_SAX is enables raw data to be specifically compared to the minimized representation and, at the same time, ensures reduced limits to Euclidean distance. It can be utilized to generate quicker, more precise algorithms for similarity searches, which improves the preciseness of time series representation through enabling better tightness of the lower bound.

AB - The operation of time series analysis to effectively manage the large amounts of data with high dimensional became an important research problem. Choose effective and scalable algorithms for appropriate representation of data is another challenge. A lot of high-level representations of the time series have been proposed for data extraction, such as spectral transfers, wavelets, piecewise polynomial, symbolic models, etc. One of the methods is Piecewise Aggregate Approximation (PAA) which minimizes dimensionality by the mean values of equal-sized frames, but this focus on mean value takes into consideration only the central tendency and not the dispersion present in each segment, which may lead to some important patterns being missed in some time series data sets. We propose method based on Time-Weighted Average for Symbolic Aggregate approximation method (TWA_SAX) compare its performance with some current methods. TWA_SAX is enables raw data to be specifically compared to the minimized representation and, at the same time, ensures reduced limits to Euclidean distance. It can be utilized to generate quicker, more precise algorithms for similarity searches, which improves the preciseness of time series representation through enabling better tightness of the lower bound.

KW - Dimensionality reduction

KW - Discretize

KW - Piecewise aggregate approximation

KW - Symbolic

KW - Time series

UR - http://www.scopus.com/inward/record.url?scp=84949742766&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84949742766&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:84949742766

VL - 7

SP - 1

EP - 15

JO - International Journal of Advances in Soft Computing and its Applications

JF - International Journal of Advances in Soft Computing and its Applications

SN - 2074-8523

IS - 3

ER -