A social network-based inference model for validating customer profile data

Sung Hyuk Park, Soon Young Huh, Wonseok Oh, Sang Han

Research output: Contribution to journalArticle

47 Citations (Scopus)

Abstract

Drawing from the social and relational perspectives, this study offers an innovative conceptualization and operational approach regarding the validation of self-reported customer demographic data, which has become an essential corporate asset for harnessing business intelligence. Specifically, based on social network and homophily paradigms in which individuals have a natural tendency to associate and interact frequently with others with similar characteristics, we constructed a relational inference model to determine the accuracy of self-administered consumer profiles. In addition, to further enhance the reliability of our model's prediction capability, we employed the entropy mechanism that minimizes potential biases that may arise from a simple probabilistic approach. To empirically validate the accuracy of our inference framework, we obtained and analyzed over 20 million actual call transactions supplied by one of the largest global telecommunication service providers. The results suggest that our social network-based inference model consistently outperforms other competing mechanisms (e.g., weighted average and simple relational classifier) regardless of the criteria choice (e.g., number of call receivers, call duration, and call frequency), with an accuracy rate of approximately 93 percent. Finally, to confirm the generalizability of our findings, we conducted simulation experiments to validate the robustness of the results in response to variations in parameter values and increases in potential noise in the data. We discuss several implications related to business intelligence for both research and practice, and offer new directions for future studies.

Original languageEnglish (US)
Pages (from-to)1217-1238
Number of pages22
JournalMIS Quarterly: Management Information Systems
Volume36
Issue number4
StatePublished - Dec 2012
Externally publishedYes

Fingerprint

Competitive intelligence
Telecommunication services
Classifiers
Entropy
Social networks
Inference
Experiments
Business intelligence
Simulation experiment
Demographics
Prediction model
Service provider
Homophily
Assets
Robustness
Paradigm
Classifier
Conceptualization
Generalizability

Keywords

  • Business intelligence
  • Customer profile
  • Data quality
  • Inference model
  • Query processing system
  • Simulation experiment
  • Social network

ASJC Scopus subject areas

  • Computer Science Applications
  • Information Systems
  • Information Systems and Management
  • Management Information Systems

Cite this

A social network-based inference model for validating customer profile data. / Park, Sung Hyuk; Huh, Soon Young; Oh, Wonseok; Han, Sang.

In: MIS Quarterly: Management Information Systems, Vol. 36, No. 4, 12.2012, p. 1217-1238.

Research output: Contribution to journalArticle

Park, Sung Hyuk ; Huh, Soon Young ; Oh, Wonseok ; Han, Sang. / A social network-based inference model for validating customer profile data. In: MIS Quarterly: Management Information Systems. 2012 ; Vol. 36, No. 4. pp. 1217-1238.
@article{d5a9e9e4a27c4d53abf526b9051c437c,
title = "A social network-based inference model for validating customer profile data",
abstract = "Drawing from the social and relational perspectives, this study offers an innovative conceptualization and operational approach regarding the validation of self-reported customer demographic data, which has become an essential corporate asset for harnessing business intelligence. Specifically, based on social network and homophily paradigms in which individuals have a natural tendency to associate and interact frequently with others with similar characteristics, we constructed a relational inference model to determine the accuracy of self-administered consumer profiles. In addition, to further enhance the reliability of our model's prediction capability, we employed the entropy mechanism that minimizes potential biases that may arise from a simple probabilistic approach. To empirically validate the accuracy of our inference framework, we obtained and analyzed over 20 million actual call transactions supplied by one of the largest global telecommunication service providers. The results suggest that our social network-based inference model consistently outperforms other competing mechanisms (e.g., weighted average and simple relational classifier) regardless of the criteria choice (e.g., number of call receivers, call duration, and call frequency), with an accuracy rate of approximately 93 percent. Finally, to confirm the generalizability of our findings, we conducted simulation experiments to validate the robustness of the results in response to variations in parameter values and increases in potential noise in the data. We discuss several implications related to business intelligence for both research and practice, and offer new directions for future studies.",
keywords = "Business intelligence, Customer profile, Data quality, Inference model, Query processing system, Simulation experiment, Social network",
author = "Park, {Sung Hyuk} and Huh, {Soon Young} and Wonseok Oh and Sang Han",
year = "2012",
month = "12",
language = "English (US)",
volume = "36",
pages = "1217--1238",
journal = "MIS Quarterly: Management Information Systems",
issn = "0276-7783",
publisher = "Management Information Systems Research Center",
number = "4",

}

TY - JOUR

T1 - A social network-based inference model for validating customer profile data

AU - Park, Sung Hyuk

AU - Huh, Soon Young

AU - Oh, Wonseok

AU - Han, Sang

PY - 2012/12

Y1 - 2012/12

N2 - Drawing from the social and relational perspectives, this study offers an innovative conceptualization and operational approach regarding the validation of self-reported customer demographic data, which has become an essential corporate asset for harnessing business intelligence. Specifically, based on social network and homophily paradigms in which individuals have a natural tendency to associate and interact frequently with others with similar characteristics, we constructed a relational inference model to determine the accuracy of self-administered consumer profiles. In addition, to further enhance the reliability of our model's prediction capability, we employed the entropy mechanism that minimizes potential biases that may arise from a simple probabilistic approach. To empirically validate the accuracy of our inference framework, we obtained and analyzed over 20 million actual call transactions supplied by one of the largest global telecommunication service providers. The results suggest that our social network-based inference model consistently outperforms other competing mechanisms (e.g., weighted average and simple relational classifier) regardless of the criteria choice (e.g., number of call receivers, call duration, and call frequency), with an accuracy rate of approximately 93 percent. Finally, to confirm the generalizability of our findings, we conducted simulation experiments to validate the robustness of the results in response to variations in parameter values and increases in potential noise in the data. We discuss several implications related to business intelligence for both research and practice, and offer new directions for future studies.

AB - Drawing from the social and relational perspectives, this study offers an innovative conceptualization and operational approach regarding the validation of self-reported customer demographic data, which has become an essential corporate asset for harnessing business intelligence. Specifically, based on social network and homophily paradigms in which individuals have a natural tendency to associate and interact frequently with others with similar characteristics, we constructed a relational inference model to determine the accuracy of self-administered consumer profiles. In addition, to further enhance the reliability of our model's prediction capability, we employed the entropy mechanism that minimizes potential biases that may arise from a simple probabilistic approach. To empirically validate the accuracy of our inference framework, we obtained and analyzed over 20 million actual call transactions supplied by one of the largest global telecommunication service providers. The results suggest that our social network-based inference model consistently outperforms other competing mechanisms (e.g., weighted average and simple relational classifier) regardless of the criteria choice (e.g., number of call receivers, call duration, and call frequency), with an accuracy rate of approximately 93 percent. Finally, to confirm the generalizability of our findings, we conducted simulation experiments to validate the robustness of the results in response to variations in parameter values and increases in potential noise in the data. We discuss several implications related to business intelligence for both research and practice, and offer new directions for future studies.

KW - Business intelligence

KW - Customer profile

KW - Data quality

KW - Inference model

KW - Query processing system

KW - Simulation experiment

KW - Social network

UR - http://www.scopus.com/inward/record.url?scp=84868007552&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84868007552&partnerID=8YFLogxK

M3 - Article

VL - 36

SP - 1217

EP - 1238

JO - MIS Quarterly: Management Information Systems

JF - MIS Quarterly: Management Information Systems

SN - 0276-7783

IS - 4

ER -