Research On The Key Causes Of Safety Accidents In Housing Municipal Engineering Based On Machine Learning

Construction industry is one of the three major industries with high accidents in China.The high frequency of accidents has not only caused great losses to the lives and property of our people,but also is an unstable factor hindering social development.Therefore,how to effectively identify the types of fatal accidents and analyze the key factors leading to accidents from fatal accident cases has gradually become the focus of research at home and abroad.The machine learning method is currently a relatively efficient data analysis method,which can fully exploit the potential unsafe factors in the massive accident cases.This study takes the recent 5 years of safety accident reports in housing municipal engineering in China as the research object,predict the four accident types and conduct a study of the key causes of the accidents.Firstly,combing,summarizing and summarizing 306 safety accident reports in housing municipal engineering in China from 2017 to 2021.After combining foreign research results,a total of 32 accident causes suitable for Chinese construction industry were extracted and the data set needed for this study was established.Secondly,three machine learning models are established,including Decision Tree(DT),Random Forest(RF)and Gradient Boosting Decision Tree(GBDT),to predict the accident types of falling,collapse,lifting injury and object striking accidents.Through the comparison and analysis of the prediction results,it is determined that the GBDT model is the optimal prediction model.Finally,calculate the permutation importance of the causes of each type of accident.Extract the top five important causes of each type of accident and then establish 625 combinations of causes.The optimal prediction model was used to predict different combination of causes.Extract the combinations of causes with the best prediction effect and establish the key causes map of safety accidents in housing municipal engineering.From the perspective of operator management,mechanical equipment management,safety management and control platform,effective management strategies are put forward for the production safety of housing municipal engineering.The key causes of accidents extracted in this study are helpful to guide the project participants to mark and warn the key fatal factors that may trigger fatal accidents,so that inspections and interventions can be implemented in a more targeted manner.