Graph Structure Learning from Unlabeled Data for Early Outbreak Detection

Sriram Somanchi, Daniel Neill

Research output: Contribution to journalArticle

Abstract

Processes such as disease propagation and information diffusion often spread over some latent network structure that must be learned from observation. Given a set of unlabeled training examples representing occurrences of an event type of interest (such as a disease outbreak), the authors aim to learn a graph structure that can be used to accurately detect future events of that type. They propose a novel framework for learning graph structure from unlabeled data by comparing the most anomalous subsets detected with and without the graph constraints. Their framework uses the mean normalized log-likelihood ratio score to measure the quality of a graph structure, and it efficiently searches for the highest-scoring graph structure. Using simulated disease outbreaks injected into real-world Emergency Department data from Allegheny County, the authors show that their method learns a structure similar to the true underlying graph, but enables faster and more accurate detection.

Original languageEnglish (US)
Article number7887641
Pages (from-to)80-84
Number of pages5
JournalIEEE Intelligent Systems
Volume32
Issue number2
DOIs
StatePublished - Mar 1 2017

Keywords

  • artificial intelligence
  • disease surveillance
  • event detection
  • graph learning
  • intelligent systems
  • spatial scan statistic

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Artificial Intelligence

Cite this

Graph Structure Learning from Unlabeled Data for Early Outbreak Detection. / Somanchi, Sriram; Neill, Daniel.

In: IEEE Intelligent Systems, Vol. 32, No. 2, 7887641, 01.03.2017, p. 80-84.

Research output: Contribution to journalArticle

@article{b90f3ed5d3d14131aae34386ea7a3e3f,
title = "Graph Structure Learning from Unlabeled Data for Early Outbreak Detection",
abstract = "Processes such as disease propagation and information diffusion often spread over some latent network structure that must be learned from observation. Given a set of unlabeled training examples representing occurrences of an event type of interest (such as a disease outbreak), the authors aim to learn a graph structure that can be used to accurately detect future events of that type. They propose a novel framework for learning graph structure from unlabeled data by comparing the most anomalous subsets detected with and without the graph constraints. Their framework uses the mean normalized log-likelihood ratio score to measure the quality of a graph structure, and it efficiently searches for the highest-scoring graph structure. Using simulated disease outbreaks injected into real-world Emergency Department data from Allegheny County, the authors show that their method learns a structure similar to the true underlying graph, but enables faster and more accurate detection.",
keywords = "artificial intelligence, disease surveillance, event detection, graph learning, intelligent systems, spatial scan statistic",
author = "Sriram Somanchi and Daniel Neill",
year = "2017",
month = "3",
day = "1",
doi = "10.1109/MIS.2017.25",
language = "English (US)",
volume = "32",
pages = "80--84",
journal = "IEEE Intelligent Systems",
issn = "1541-1672",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "2",

}

TY - JOUR

T1 - Graph Structure Learning from Unlabeled Data for Early Outbreak Detection

AU - Somanchi, Sriram

AU - Neill, Daniel

PY - 2017/3/1

Y1 - 2017/3/1

N2 - Processes such as disease propagation and information diffusion often spread over some latent network structure that must be learned from observation. Given a set of unlabeled training examples representing occurrences of an event type of interest (such as a disease outbreak), the authors aim to learn a graph structure that can be used to accurately detect future events of that type. They propose a novel framework for learning graph structure from unlabeled data by comparing the most anomalous subsets detected with and without the graph constraints. Their framework uses the mean normalized log-likelihood ratio score to measure the quality of a graph structure, and it efficiently searches for the highest-scoring graph structure. Using simulated disease outbreaks injected into real-world Emergency Department data from Allegheny County, the authors show that their method learns a structure similar to the true underlying graph, but enables faster and more accurate detection.

AB - Processes such as disease propagation and information diffusion often spread over some latent network structure that must be learned from observation. Given a set of unlabeled training examples representing occurrences of an event type of interest (such as a disease outbreak), the authors aim to learn a graph structure that can be used to accurately detect future events of that type. They propose a novel framework for learning graph structure from unlabeled data by comparing the most anomalous subsets detected with and without the graph constraints. Their framework uses the mean normalized log-likelihood ratio score to measure the quality of a graph structure, and it efficiently searches for the highest-scoring graph structure. Using simulated disease outbreaks injected into real-world Emergency Department data from Allegheny County, the authors show that their method learns a structure similar to the true underlying graph, but enables faster and more accurate detection.

KW - artificial intelligence

KW - disease surveillance

KW - event detection

KW - graph learning

KW - intelligent systems

KW - spatial scan statistic

UR - http://www.scopus.com/inward/record.url?scp=85017157738&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85017157738&partnerID=8YFLogxK

U2 - 10.1109/MIS.2017.25

DO - 10.1109/MIS.2017.25

M3 - Article

VL - 32

SP - 80

EP - 84

JO - IEEE Intelligent Systems

JF - IEEE Intelligent Systems

SN - 1541-1672

IS - 2

M1 - 7887641

ER -