Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
Eine Plattform für die Wissenschaft: Bauingenieurwesen, Architektur und Urbanistik
The Analysis of Construction Workplace Incident Fatalities Using Text Mining and Classification Techniques
https://orcid.org/http://orcid.org/0000-0003-3256-3310
https://orcid.org/http://orcid.org/0000-0002-1262-1440
https://orcid.org/http://orcid.org/0000-0002-1115-7842
https://orcid.org/http://orcid.org/0000-0002-3835-3931
https://orcid.org/http://orcid.org/0000-0002-2790-4347
https://orcid.org/http://orcid.org/0000-0001-9170-9557
The construction industry in Canada has been witnessing an increased number of workplace fatalities. In fact, a total of 52 fatalities have been observed in Alberta’s construction industry in 2020 according to Alberta Occupational Health and Safety (OHS). Alberta OHS works on investigating such incidents and prepares the corresponding investigation reports. However, these are presented in a textual format making it challenging to extract further insights. Additionally, these reports present knowledge on the causes of such fatalities which can be potentially leveraged in identifying the necessary preventive measures for other similar projects. As such, the objective of this research work is two-fold. First, it aims to obtain insights related to workplace incident fatalities in construction projects by analyzing 116 reports between 2004 and 2022. This will be supported by the use of text mining analysis as a preliminary step to transform the data into a structured format. Secondly, this work aims to develop and compare different machine learning-based classifiers to identify the probability of occurrence of a specified type of a certain fatality incident at the construction site given a potential-incident’s information. Text mining results of 100 incident investigation reports showed that fall types of fatalities are the most common. Additionally, the performance of Decision Trees, Random Forest, Gradient Boosted Tree, and the Zero Classifier were assessed in classifying the type of fatality. The Random Forest classifier was selected for classifying Crush and Fall types of fatality incidents as it gave the highest F1 scores of 90.5% and 88.3%, respectively. Safety personnel can leverage the developed classifier to integrate a data-driven approach in identifying the necessary preventive measures to mitigate fatal incidents on construction sites.
The Analysis of Construction Workplace Incident Fatalities Using Text Mining and Classification Techniques
https://orcid.org/http://orcid.org/0000-0003-3256-3310
https://orcid.org/http://orcid.org/0000-0002-1262-1440
https://orcid.org/http://orcid.org/0000-0002-1115-7842
https://orcid.org/http://orcid.org/0000-0002-3835-3931
https://orcid.org/http://orcid.org/0000-0002-2790-4347
https://orcid.org/http://orcid.org/0000-0001-9170-9557
The construction industry in Canada has been witnessing an increased number of workplace fatalities. In fact, a total of 52 fatalities have been observed in Alberta’s construction industry in 2020 according to Alberta Occupational Health and Safety (OHS). Alberta OHS works on investigating such incidents and prepares the corresponding investigation reports. However, these are presented in a textual format making it challenging to extract further insights. Additionally, these reports present knowledge on the causes of such fatalities which can be potentially leveraged in identifying the necessary preventive measures for other similar projects. As such, the objective of this research work is two-fold. First, it aims to obtain insights related to workplace incident fatalities in construction projects by analyzing 116 reports between 2004 and 2022. This will be supported by the use of text mining analysis as a preliminary step to transform the data into a structured format. Secondly, this work aims to develop and compare different machine learning-based classifiers to identify the probability of occurrence of a specified type of a certain fatality incident at the construction site given a potential-incident’s information. Text mining results of 100 incident investigation reports showed that fall types of fatalities are the most common. Additionally, the performance of Decision Trees, Random Forest, Gradient Boosted Tree, and the Zero Classifier were assessed in classifying the type of fatality. The Random Forest classifier was selected for classifying Crush and Fall types of fatality incidents as it gave the highest F1 scores of 90.5% and 88.3%, respectively. Safety personnel can leverage the developed classifier to integrate a data-driven approach in identifying the necessary preventive measures to mitigate fatal incidents on construction sites.
The Analysis of Construction Workplace Incident Fatalities Using Text Mining and Classification Techniques
https://orcid.org/http://orcid.org/0000-0003-3256-3310
https://orcid.org/http://orcid.org/0000-0002-1262-1440
https://orcid.org/http://orcid.org/0000-0002-1115-7842
https://orcid.org/http://orcid.org/0000-0002-3835-3931
https://orcid.org/http://orcid.org/0000-0002-2790-4347
https://orcid.org/http://orcid.org/0000-0001-9170-9557
The construction industry in Canada has been witnessing an increased number of workplace fatalities. In fact, a total of 52 fatalities have been observed in Alberta’s construction industry in 2020 according to Alberta Occupational Health and Safety (OHS). Alberta OHS works on investigating such incidents and prepares the corresponding investigation reports. However, these are presented in a textual format making it challenging to extract further insights. Additionally, these reports present knowledge on the causes of such fatalities which can be potentially leveraged in identifying the necessary preventive measures for other similar projects. As such, the objective of this research work is two-fold. First, it aims to obtain insights related to workplace incident fatalities in construction projects by analyzing 116 reports between 2004 and 2022. This will be supported by the use of text mining analysis as a preliminary step to transform the data into a structured format. Secondly, this work aims to develop and compare different machine learning-based classifiers to identify the probability of occurrence of a specified type of a certain fatality incident at the construction site given a potential-incident’s information. Text mining results of 100 incident investigation reports showed that fall types of fatalities are the most common. Additionally, the performance of Decision Trees, Random Forest, Gradient Boosted Tree, and the Zero Classifier were assessed in classifying the type of fatality. The Random Forest classifier was selected for classifying Crush and Fall types of fatality incidents as it gave the highest F1 scores of 90.5% and 88.3%, respectively. Safety personnel can leverage the developed classifier to integrate a data-driven approach in identifying the necessary preventive measures to mitigate fatal incidents on construction sites.
The Analysis of Construction Workplace Incident Fatalities Using Text Mining and Classification Techniques
Lecture Notes in Civil Engineering
Desjardins, Serge (Herausgeber:in) / Poitras, Gérard J. (Herausgeber:in) / Nik-Bakht, Mazdak (Herausgeber:in) / Assaf, Sena (Autor:in) / Liu, Kexin (Autor:in) / Mao, Zeyu (Autor:in) / Saleh, Amira (Autor:in) / Abellanosa, Abbey (Autor:in) / Mohamed, Yasser (Autor:in)
Canadian Society of Civil Engineering Annual Conference ; 2023 ; Moncton, NB, Canada
Proceedings of the Canadian Society for Civil Engineering Annual Conference 2023, Volume 5 ; Kapitel: 12 ; 157-171
18.12.2024
15 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Fatalities in construction industry
Emerald Group Publishing | 1999
Automated Classification of Construction Claim Documents Using Text Mining
| Springer Verlag | 2024
Construction site accident analysis using text mining and natural language processing techniques
| British Library Online Contents | 2019
Text mining for occupant perspectives on the physical workplace
| Online Contents | 2011
Text mining for occupant perspectives on the physical workplace
| Taylor & Francis Verlag | 2011