Missing Data Imputation and the Inductive Modelling

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F07%3A03132900" target="_blank" >RIV/68407700:21230/07:03132900 - isvavai.cz</a>

Result on the web

—

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Missing Data Imputation and the Inductive Modelling

Original language description

Missing data is a big problem in simulation for data mining and data analysis. Real world applications often contains missing data. Many data-mining methods is unable to create models from data which contains missing values. Traditional approach is to delete vectors with missing data. Unfortunately, this approach may lead to decreased accuracy of the models and in the worst case all data in dataset may be deleted. For this reason many different imputation techniques were developed and some are widely used. In this paper, we present a comparison of several well-known techniques for missing data imputation. Presented techniques includes imputation of mean value, zero, value from nearest input vector and few others. In this paper we show which techniquesare the best in estimation of missing values. To test imputation methods we used several different datasets. We compare the imputation methods in two ways. The first is to compare imputed data with original data.

Czech name

Nahrazování chybějících dat a induktivní modelování

Czech description

Missing data is a big problem in simulation for data mining and data analysis. Real world applications often contains missing data. Many data-mining methods is unable to create models from data which contains missing values. Traditional approach is to delete vectors with missing data. Unfortunately, this approach may lead to decreased accuracy of the models and in the worst case all data in dataset may be deleted. For this reason many different imputation techniques were developed and some are widely used. In this paper, we present a comparison of several well-known techniques for missing data imputation. Presented techniques includes imputation of mean value, zero, value from nearest input vector and few others. In this paper we show which techniquesare the best in estimation of missing values. To test imputation methods we used several different datasets. We compare the imputation methods in two ways. The first is to compare imputed data with original data.

Type

D - Article in proceedings

CEP classification

IN - Informatics

OECD FORD branch

—

Project

<a href="/en/project/KJB201210701" target="_blank" >KJB201210701: Automated Knowledge Extraction</a><br>

Continuities

Z - Vyzkumny zamer (s odkazem do CEZ)

Publication year

2007

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

Proceedings of the 6th EUROSIM Congress on Modelling and Simulation

ISBN

978-3-901608-32-2

ISSN

—

e-ISSN

—

Number of pages

8

Pages from-to

—

Publisher name

ARGESIM

Place of publication

Vienna

Event location

Ljubljana

Event date

Sep 9, 2007

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—