Data mining on fire records of New South Wales, Sydney

Eric Wai-ming Lee; Guan-Heng Yeoh; Morgan Cook; Chris Lewis

doi:10.1016/j.proeng.2014.04.047

Data mining on fire records of New South Wales, Sydney

Eric Wai-ming Lee^*, Guan-Heng Yeoh, Morgan Cook, Chris Lewis

^*Corresponding author for this work

Research output: Journal Publications and Reviews › RGC 21 - Publication in refereed journal › peer-review

2 Citations (Scopus)

64 Downloads (CityUHK Scholars)

Abstract

This study gathered fire records from the Fire and Rescue New South Wales (F&RNSW) for investigating the most relevant event to the fire accident. Support vector machine was adopted to mimic the correlation between the information of the buiding and occupants and the occurrence of fire accident. The percentage of correct prediction is 65% which is considered reasonable since noise is expected to be embedded in the data of the fire records. Bayesian approach was also adopted to anayze the relevancies of the binary input parameters to the fire occurrence. Monte Carlo simulation was conducted. The result shows that the Special-Risk-Buiding and Smokers are the two parameters most relevant to the occurrence of fire accident. © 2014 Published by Elsevier Ltd.

Original language	English
Pages (from-to)	328-332
Journal	Procedia Engineering
Volume	71
Online published	16 May 2014
DOIs	https://doi.org/10.1016/j.proeng.2014.04.047
Publication status	Published - 2014
Event	2013 International Conference on Performance-Based Fire and Fire Protection Engineering, ICPFFPE 2013 - Wuhan, China Duration: 16 Nov 2013 → 17 Nov 2013

Research Keywords

Bayesian theorem
Fire records
Monte carlo simulation
Support vector machine

Publisher's Copyright Statement

This full text is made available under CC-BY-NC-ND 3.0. https://creativecommons.org/licenses/by-nc-nd/3.0/

Access Output

10.1016/j.proeng.2014.04.047

5956047.pdfFinal Published version, 356 KBLicence: CC BY-NC-ND

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

@article{f3bb54655d744de8bdc8f72c3857df24,

title = "Data mining on fire records of New South Wales, Sydney",

abstract = "This study gathered fire records from the Fire and Rescue New South Wales (F&RNSW) for investigating the most relevant event to the fire accident. Support vector machine was adopted to mimic the correlation between the information of the buiding and occupants and the occurrence of fire accident. The percentage of correct prediction is 65% which is considered reasonable since noise is expected to be embedded in the data of the fire records. Bayesian approach was also adopted to anayze the relevancies of the binary input parameters to the fire occurrence. Monte Carlo simulation was conducted. The result shows that the Special-Risk-Buiding and Smokers are the two parameters most relevant to the occurrence of fire accident. {\textcopyright} 2014 Published by Elsevier Ltd.",

keywords = "Bayesian theorem, Fire records, Monte carlo simulation, Support vector machine",

author = "Lee, {Eric Wai-ming} and Guan-Heng Yeoh and Morgan Cook and Chris Lewis",

year = "2014",

doi = "10.1016/j.proeng.2014.04.047",

language = "English",

volume = "71",

pages = "328--332",

journal = "Procedia Engineering",

issn = "1877-7058",

publisher = "Elsevier B.V.",

note = "2013 International Conference on Performance-Based Fire and Fire Protection Engineering, ICPFFPE 2013 ; Conference date: 16-11-2013 Through 17-11-2013",

}

TY - JOUR

T1 - Data mining on fire records of New South Wales, Sydney

AU - Lee, Eric Wai-ming

AU - Yeoh, Guan-Heng

AU - Cook, Morgan

AU - Lewis, Chris

PY - 2014

Y1 - 2014

N2 - This study gathered fire records from the Fire and Rescue New South Wales (F&RNSW) for investigating the most relevant event to the fire accident. Support vector machine was adopted to mimic the correlation between the information of the buiding and occupants and the occurrence of fire accident. The percentage of correct prediction is 65% which is considered reasonable since noise is expected to be embedded in the data of the fire records. Bayesian approach was also adopted to anayze the relevancies of the binary input parameters to the fire occurrence. Monte Carlo simulation was conducted. The result shows that the Special-Risk-Buiding and Smokers are the two parameters most relevant to the occurrence of fire accident. © 2014 Published by Elsevier Ltd.

AB - This study gathered fire records from the Fire and Rescue New South Wales (F&RNSW) for investigating the most relevant event to the fire accident. Support vector machine was adopted to mimic the correlation between the information of the buiding and occupants and the occurrence of fire accident. The percentage of correct prediction is 65% which is considered reasonable since noise is expected to be embedded in the data of the fire records. Bayesian approach was also adopted to anayze the relevancies of the binary input parameters to the fire occurrence. Monte Carlo simulation was conducted. The result shows that the Special-Risk-Buiding and Smokers are the two parameters most relevant to the occurrence of fire accident. © 2014 Published by Elsevier Ltd.

KW - Bayesian theorem

KW - Fire records

KW - Monte carlo simulation

KW - Support vector machine

UR - http://www.scopus.com/inward/record.url?scp=84901827697&partnerID=8YFLogxK

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-84901827697&origin=recordpage

U2 - 10.1016/j.proeng.2014.04.047

DO - 10.1016/j.proeng.2014.04.047

M3 - RGC 21 - Publication in refereed journal

SN - 1877-7058

VL - 71

SP - 328

EP - 332

JO - Procedia Engineering

JF - Procedia Engineering

T2 - 2013 International Conference on Performance-Based Fire and Fire Protection Engineering, ICPFFPE 2013

Y2 - 16 November 2013 through 17 November 2013

ER -

Data mining on fire records of New South Wales, Sydney

Abstract

Research Keywords

Publisher's Copyright Statement

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this