Incomplete information imputation in limited data environments with application to disaster response

Kezban Yagci Sokat*, Irina S. Dolinskaya, Karen Renee Smilowitz, Ryan Bank

*Corresponding author for this work

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Following a major disaster, a field operations manager needs to deploy relief activities within the affected region. State-of-the-art humanitarian logistics models have been developed over the past decades to assist relief operations. However, while many models assume availability of information on infrastructure status, this is typically not the case in practice. Often, only partial information about infrastructure status is known. Utilizing the similarities in the known data via attributes, we develop a framework to impute incomplete information in limited data environments. We present an application of this framework to a past disaster, the 2010 Haiti earthquake. We build an ArcGIS model to automate the data collection and processing efforts to the extent possible. The study explores the impact of missing data, dispersion of missing data and imputation techniques used in approximating the incomplete information. Our results suggest that lower granularity yields better estimates of the unknown information above a threshold. We also develop publicly available test cases for the broader community.

Original languageEnglish (US)
Pages (from-to)466-485
Number of pages20
JournalEuropean Journal of Operational Research
Volume269
Issue number2
DOIs
StatePublished - Sep 1 2018

Fingerprint

Imputation
Incomplete Information
Disaster
Missing Data
Disasters
Infrastructure
Logistic Model
Partial Information
Granularity
Earthquake
Availability
Attribute
Unknown
Logistics
Earthquakes
Managers
Model
Estimate
Processing
Framework

Keywords

  • 2010 Haiti earthquake
  • Humanitarian logistics
  • Imputation
  • Incomplete information
  • Limited data

ASJC Scopus subject areas

  • Modeling and Simulation
  • Management Science and Operations Research
  • Information Systems and Management

Cite this

@article{401fa1176bf6422183da7045017df545,
title = "Incomplete information imputation in limited data environments with application to disaster response",
abstract = "Following a major disaster, a field operations manager needs to deploy relief activities within the affected region. State-of-the-art humanitarian logistics models have been developed over the past decades to assist relief operations. However, while many models assume availability of information on infrastructure status, this is typically not the case in practice. Often, only partial information about infrastructure status is known. Utilizing the similarities in the known data via attributes, we develop a framework to impute incomplete information in limited data environments. We present an application of this framework to a past disaster, the 2010 Haiti earthquake. We build an ArcGIS model to automate the data collection and processing efforts to the extent possible. The study explores the impact of missing data, dispersion of missing data and imputation techniques used in approximating the incomplete information. Our results suggest that lower granularity yields better estimates of the unknown information above a threshold. We also develop publicly available test cases for the broader community.",
keywords = "2010 Haiti earthquake, Humanitarian logistics, Imputation, Incomplete information, Limited data",
author = "{Yagci Sokat}, Kezban and Dolinskaya, {Irina S.} and Smilowitz, {Karen Renee} and Ryan Bank",
year = "2018",
month = "9",
day = "1",
doi = "10.1016/j.ejor.2018.02.016",
language = "English (US)",
volume = "269",
pages = "466--485",
journal = "European Journal of Operational Research",
issn = "0377-2217",
publisher = "Elsevier",
number = "2",

}

Incomplete information imputation in limited data environments with application to disaster response. / Yagci Sokat, Kezban; Dolinskaya, Irina S.; Smilowitz, Karen Renee; Bank, Ryan.

In: European Journal of Operational Research, Vol. 269, No. 2, 01.09.2018, p. 466-485.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Incomplete information imputation in limited data environments with application to disaster response

AU - Yagci Sokat, Kezban

AU - Dolinskaya, Irina S.

AU - Smilowitz, Karen Renee

AU - Bank, Ryan

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Following a major disaster, a field operations manager needs to deploy relief activities within the affected region. State-of-the-art humanitarian logistics models have been developed over the past decades to assist relief operations. However, while many models assume availability of information on infrastructure status, this is typically not the case in practice. Often, only partial information about infrastructure status is known. Utilizing the similarities in the known data via attributes, we develop a framework to impute incomplete information in limited data environments. We present an application of this framework to a past disaster, the 2010 Haiti earthquake. We build an ArcGIS model to automate the data collection and processing efforts to the extent possible. The study explores the impact of missing data, dispersion of missing data and imputation techniques used in approximating the incomplete information. Our results suggest that lower granularity yields better estimates of the unknown information above a threshold. We also develop publicly available test cases for the broader community.

AB - Following a major disaster, a field operations manager needs to deploy relief activities within the affected region. State-of-the-art humanitarian logistics models have been developed over the past decades to assist relief operations. However, while many models assume availability of information on infrastructure status, this is typically not the case in practice. Often, only partial information about infrastructure status is known. Utilizing the similarities in the known data via attributes, we develop a framework to impute incomplete information in limited data environments. We present an application of this framework to a past disaster, the 2010 Haiti earthquake. We build an ArcGIS model to automate the data collection and processing efforts to the extent possible. The study explores the impact of missing data, dispersion of missing data and imputation techniques used in approximating the incomplete information. Our results suggest that lower granularity yields better estimates of the unknown information above a threshold. We also develop publicly available test cases for the broader community.

KW - 2010 Haiti earthquake

KW - Humanitarian logistics

KW - Imputation

KW - Incomplete information

KW - Limited data

UR - http://www.scopus.com/inward/record.url?scp=85042517121&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85042517121&partnerID=8YFLogxK

U2 - 10.1016/j.ejor.2018.02.016

DO - 10.1016/j.ejor.2018.02.016

M3 - Article

VL - 269

SP - 466

EP - 485

JO - European Journal of Operational Research

JF - European Journal of Operational Research

SN - 0377-2217

IS - 2

ER -