Text Mining Based Analysis For Construction Engineering Accident Reports

Safety management is one of the most important tasks in construction management.At present,the progress of national strength has led to the rapid development of the construction industry,but the accidents and deaths caused by construction are also increasing.In order to reduce the occurrence of construction accidents and better protect the physical and mental health of construction workers,it is necessary to strengthen the study of construction accident reports and sum up the experience.In the past,the summary of construction safety risk is mainly through the relevant personnel from the case or related areas of knowledge to manually sum up experience and form a set of safety management.In order to explore the potential rules of accident reporting,form useful safety management information to provide advice for safety management,more advanced tools or analysis methods are needed.Based on the theory and technology of natural language processing,this paper compares the advantages and disadvantages of each machine learning method in dealing with the classification task of construction accident report.The model can extract semantic features from construction accident reports and automatically classify them into predefined categories.The results show that the classification method based on C-BiLSTM has the best classification effect in the case of less manual intervention and complex feature engineering.In order to make further analysis of accident report,this paper uses association rules technology of data mining to further study the casualties caused by three major accidents,and puts forward corresponding suggestions.Through verification,the results of classifying construction accident texts by C-BiLSTM model are basically consistent with the actual situation,and the analysis results of related reports by association rules are also consistent with the manual experience.This provides a reliable analysis framework for the follow-up safety management analysis,and is conducive to the scientific a nd intelligent development of safety management.