A neural network-based framework for the reconstruction of incomplete data sets

Iffat A. Gheyas, Leslie S. Smith

Research output: Contribution to journalArticle

47 Citations (Scopus)

Abstract

The treatment of incomplete data is an important step in the pre-processing of data. We propose a novel nonparametric algorithm Generalized regression neural network Ensemble for Multiple Imputation (GEMI). We also developed a single imputation (SI) version of this approach-GESI. We compare our algorithms with 25 popular missing data imputation algorithms on 98 real-world and synthetic datasets for various percentage of missing values. The effectiveness of the algorithms is evaluated in terms of (i) the accuracy of output classification: three classifiers (a generalized regression neural network, a multilayer perceptron and a logistic regression technique) are separately trained and tested on the dataset imputed with each imputation algorithm, (ii) interval analysis with missing observations and (iii) point estimation accuracy of the missing value imputation. GEMI outperformed GESI and all the conventional imputation algorithms in terms of all three criteria considered.
Original languageEnglish
Pages (from-to)3039-3065
Number of pages27
JournalNeurocomputing
Volume73
Issue number16-18
Early online date8 Aug 2010
DOIs
Publication statusPublished - Oct 2010
Externally publishedYes

Fingerprint

Neural networks
Neural Networks (Computer)
Multilayer neural networks
Datasets
Logistics
Classifiers
Logistic Models
Processing

Cite this

Gheyas, Iffat A. ; Smith, Leslie S. / A neural network-based framework for the reconstruction of incomplete data sets. In: Neurocomputing. 2010 ; Vol. 73, No. 16-18. pp. 3039-3065.
@article{633e10bf588340608529bd1c081b8c8e,
title = "A neural network-based framework for the reconstruction of incomplete data sets",
abstract = "The treatment of incomplete data is an important step in the pre-processing of data. We propose a novel nonparametric algorithm Generalized regression neural network Ensemble for Multiple Imputation (GEMI). We also developed a single imputation (SI) version of this approach-GESI. We compare our algorithms with 25 popular missing data imputation algorithms on 98 real-world and synthetic datasets for various percentage of missing values. The effectiveness of the algorithms is evaluated in terms of (i) the accuracy of output classification: three classifiers (a generalized regression neural network, a multilayer perceptron and a logistic regression technique) are separately trained and tested on the dataset imputed with each imputation algorithm, (ii) interval analysis with missing observations and (iii) point estimation accuracy of the missing value imputation. GEMI outperformed GESI and all the conventional imputation algorithms in terms of all three criteria considered.",
keywords = "Generalized regression neural networks, Imputation, Missing values, Multiple imputation, Single imputation",
author = "Gheyas, {Iffat A.} and Smith, {Leslie S.}",
year = "2010",
month = "10",
doi = "10.1016/j.neucom.2010.06.021",
language = "English",
volume = "73",
pages = "3039--3065",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "16-18",

}

A neural network-based framework for the reconstruction of incomplete data sets. / Gheyas, Iffat A.; Smith, Leslie S.

In: Neurocomputing, Vol. 73, No. 16-18, 10.2010, p. 3039-3065.

Research output: Contribution to journalArticle

TY - JOUR

T1 - A neural network-based framework for the reconstruction of incomplete data sets

AU - Gheyas, Iffat A.

AU - Smith, Leslie S.

PY - 2010/10

Y1 - 2010/10

N2 - The treatment of incomplete data is an important step in the pre-processing of data. We propose a novel nonparametric algorithm Generalized regression neural network Ensemble for Multiple Imputation (GEMI). We also developed a single imputation (SI) version of this approach-GESI. We compare our algorithms with 25 popular missing data imputation algorithms on 98 real-world and synthetic datasets for various percentage of missing values. The effectiveness of the algorithms is evaluated in terms of (i) the accuracy of output classification: three classifiers (a generalized regression neural network, a multilayer perceptron and a logistic regression technique) are separately trained and tested on the dataset imputed with each imputation algorithm, (ii) interval analysis with missing observations and (iii) point estimation accuracy of the missing value imputation. GEMI outperformed GESI and all the conventional imputation algorithms in terms of all three criteria considered.

AB - The treatment of incomplete data is an important step in the pre-processing of data. We propose a novel nonparametric algorithm Generalized regression neural network Ensemble for Multiple Imputation (GEMI). We also developed a single imputation (SI) version of this approach-GESI. We compare our algorithms with 25 popular missing data imputation algorithms on 98 real-world and synthetic datasets for various percentage of missing values. The effectiveness of the algorithms is evaluated in terms of (i) the accuracy of output classification: three classifiers (a generalized regression neural network, a multilayer perceptron and a logistic regression technique) are separately trained and tested on the dataset imputed with each imputation algorithm, (ii) interval analysis with missing observations and (iii) point estimation accuracy of the missing value imputation. GEMI outperformed GESI and all the conventional imputation algorithms in terms of all three criteria considered.

KW - Generalized regression neural networks

KW - Imputation

KW - Missing values

KW - Multiple imputation

KW - Single imputation

UR - http://www.scopus.com/inward/record.url?scp=78650175506&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2010.06.021

DO - 10.1016/j.neucom.2010.06.021

M3 - Article

VL - 73

SP - 3039

EP - 3065

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 16-18

ER -