Automatic instance selection via locality constrained sparse representation for missing value estimation

Xiaodong Feng, Sen Wu, Jaideep Srivastava, Prasanna Desikan

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Missing values in real application can significantly disturb the result of knowledge discovery, and it is thus vital to estimate this unknown data accurately. This paper focuses on applying sparse representation to improve the quality of estimation of the absent values. Firstly, a novel sparse representation scheme called locality constrained sparse representation (LCSR) is presented, introducing locality l<inf>1</inf>-norm and l<inf>2</inf>-norm regularization. Taking the advantage of sparsity, smoothness and locality structure, LCSR is capable of automatically selecting instance and avoiding overfitting. Then LCSR-based missing value estimation (LCSR-MVE) is proposed to estimate the unobserved values through the linear combination of automatically selected atoms from dictionary due to the sparsity in reconstruction coefficient vector, while three dictionary constructions are also developed respectively. The proposed LCSR-MVE is evaluated on 6 datasets from UCI and gene expression databases, compared with other instance-based missing value estimation methods. Results show that the proposed LCSR-MVE outperforms other state-of-arts methods in terms of normalized root mean squared error (NRMSE), and is not much sensitive to the dictionary size and regularization parameters.

Original languageEnglish
Pages (from-to)210-223
Number of pages14
JournalKnowledge-Based Systems
Volume85
DOIs
Publication statusPublished - 1 Sep 2015
Externally publishedYes

Fingerprint

Glossaries
Gene expression
Data mining
Atoms
Locality
Missing values

Keywords

  • Instance selection
  • Locality constrained regularization
  • Missing value estimation
  • Sparse representation

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence
  • Management Information Systems
  • Information Systems and Management

Cite this

Automatic instance selection via locality constrained sparse representation for missing value estimation. / Feng, Xiaodong; Wu, Sen; Srivastava, Jaideep; Desikan, Prasanna.

In: Knowledge-Based Systems, Vol. 85, 01.09.2015, p. 210-223.

Research output: Contribution to journalArticle

@article{ffcfd4b5b20444ab9550cce3b7c85726,
title = "Automatic instance selection via locality constrained sparse representation for missing value estimation",
abstract = "Missing values in real application can significantly disturb the result of knowledge discovery, and it is thus vital to estimate this unknown data accurately. This paper focuses on applying sparse representation to improve the quality of estimation of the absent values. Firstly, a novel sparse representation scheme called locality constrained sparse representation (LCSR) is presented, introducing locality l1-norm and l2-norm regularization. Taking the advantage of sparsity, smoothness and locality structure, LCSR is capable of automatically selecting instance and avoiding overfitting. Then LCSR-based missing value estimation (LCSR-MVE) is proposed to estimate the unobserved values through the linear combination of automatically selected atoms from dictionary due to the sparsity in reconstruction coefficient vector, while three dictionary constructions are also developed respectively. The proposed LCSR-MVE is evaluated on 6 datasets from UCI and gene expression databases, compared with other instance-based missing value estimation methods. Results show that the proposed LCSR-MVE outperforms other state-of-arts methods in terms of normalized root mean squared error (NRMSE), and is not much sensitive to the dictionary size and regularization parameters.",
keywords = "Instance selection, Locality constrained regularization, Missing value estimation, Sparse representation",
author = "Xiaodong Feng and Sen Wu and Jaideep Srivastava and Prasanna Desikan",
year = "2015",
month = "9",
day = "1",
doi = "10.1016/j.knosys.2015.05.007",
language = "English",
volume = "85",
pages = "210--223",
journal = "Knowledge-Based Systems",
issn = "0950-7051",
publisher = "Elsevier",

}

TY - JOUR

T1 - Automatic instance selection via locality constrained sparse representation for missing value estimation

AU - Feng, Xiaodong

AU - Wu, Sen

AU - Srivastava, Jaideep

AU - Desikan, Prasanna

PY - 2015/9/1

Y1 - 2015/9/1

N2 - Missing values in real application can significantly disturb the result of knowledge discovery, and it is thus vital to estimate this unknown data accurately. This paper focuses on applying sparse representation to improve the quality of estimation of the absent values. Firstly, a novel sparse representation scheme called locality constrained sparse representation (LCSR) is presented, introducing locality l1-norm and l2-norm regularization. Taking the advantage of sparsity, smoothness and locality structure, LCSR is capable of automatically selecting instance and avoiding overfitting. Then LCSR-based missing value estimation (LCSR-MVE) is proposed to estimate the unobserved values through the linear combination of automatically selected atoms from dictionary due to the sparsity in reconstruction coefficient vector, while three dictionary constructions are also developed respectively. The proposed LCSR-MVE is evaluated on 6 datasets from UCI and gene expression databases, compared with other instance-based missing value estimation methods. Results show that the proposed LCSR-MVE outperforms other state-of-arts methods in terms of normalized root mean squared error (NRMSE), and is not much sensitive to the dictionary size and regularization parameters.

AB - Missing values in real application can significantly disturb the result of knowledge discovery, and it is thus vital to estimate this unknown data accurately. This paper focuses on applying sparse representation to improve the quality of estimation of the absent values. Firstly, a novel sparse representation scheme called locality constrained sparse representation (LCSR) is presented, introducing locality l1-norm and l2-norm regularization. Taking the advantage of sparsity, smoothness and locality structure, LCSR is capable of automatically selecting instance and avoiding overfitting. Then LCSR-based missing value estimation (LCSR-MVE) is proposed to estimate the unobserved values through the linear combination of automatically selected atoms from dictionary due to the sparsity in reconstruction coefficient vector, while three dictionary constructions are also developed respectively. The proposed LCSR-MVE is evaluated on 6 datasets from UCI and gene expression databases, compared with other instance-based missing value estimation methods. Results show that the proposed LCSR-MVE outperforms other state-of-arts methods in terms of normalized root mean squared error (NRMSE), and is not much sensitive to the dictionary size and regularization parameters.

KW - Instance selection

KW - Locality constrained regularization

KW - Missing value estimation

KW - Sparse representation

UR - http://www.scopus.com/inward/record.url?scp=84937516978&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84937516978&partnerID=8YFLogxK

U2 - 10.1016/j.knosys.2015.05.007

DO - 10.1016/j.knosys.2015.05.007

M3 - Article

AN - SCOPUS:84937516978

VL - 85

SP - 210

EP - 223

JO - Knowledge-Based Systems

JF - Knowledge-Based Systems

SN - 0950-7051

ER -