건축도시공간연구소

Architecture & Urban Research Institute

pdf원문보기 에러 해결방법 바로가기



문헌홈 > 연구논문 > 상세

[원문보기시 소비되는 포인트 : 100 포인트] 미리보기 인용

한국생태환경건축학회|KIEAE Journal 2024년 4월

논문명 머신러닝 모델을 활용한 산업단지 화재 재산피해 크기 예측 / Predicting the Size of Fire Property Damage in an Industrial Complex Using a Machine Learning Model
저자명 이종호 ; 최규진 ; 박초롱 ; 이재욱 ; 손동욱
발행사 한국생태환경건축학회
수록사항 KIEAE Journal, Vol.24 No.2(통권 126호) (2024-04)
페이지 시작페이지(97) 총페이지(10)
ISSN 2288-968X
주제분류 환경및설비
주제어 기계학습; 산업단지; 화재 재산피해; 공간지리정보 ; Machine Learning; Industrial Complex; Fire Property Damage; GIS
요약2 Purpose: This study proposes a novel approach to reduce the severe damages caused by factory fires in South Korea. The current fire risk assessment system faces limitations in providing detailed evaluations for factory buildings. This research utilizes public data and machine learning to swiftly and accurately predict fire risks in factories and seeks methods to identify and manage high-risk areas within industrial complexes. Method: The research process encompasses data collection, preprocessing, model prediction, and the integration of spatial data using GIS. It leverages building information provided by the national data portal and fire scenario data set as control variables. Data preprocessing includes the simplification of categorical variables, creation of derived variables, and the conversion of string data into numeric data. The predictive outcomes are integrated with spatial data using GIS, and industrial complexes are subdivided into blocks for risk level grading. This method aims to make a practical contribution to the management and prevention of fire risks in industrial complexes. Result: This study classified and analyzed the characteristics of factory buildings in aged three industrial complexes, assessing regional differences. Utilizing the Random Forest model, fire risks were categorized into low, medium, and severe levels, and regression analysis was employed to evaluate the impact of factors on fire risk. A five-tier grading system based on GIS visualization comprehensively represents the fire risk by region, offering valuable information for fire risk management. This research contributes to the development of policies aimed at enhancing safety in industrial complexes and minimizing property loss.
소장처 한국생태환경건축학회
언어 한국어
DOI http://doi.org/10.12813/kieae.2024.24.2.097