A Correlation Analysis of Construction Site Fall Accidents Based on Text Mining

As construction falling accidents are a type of high-frequency accident in the construction industry, the analysis of their causes and the study of risk management have attracted extensive attention. However, the application of unstructured text data to explore the connections between the causes of these accidents urgently needs to be studied. This paper adopted text mining methods to analyse and process the accident reports of 557 domestic construction fall accidents from 2013 to 2019. First, the text of these accident reports was preprocessed to identify 6 types and 28 causes of high-fall accidents; subsequently, the 28 causes were divided into key causes, subcritical causes and general causes according to their importance. Then, Apriori software was used to analyse the associations among these high-fall accidents. Finally, a series of strong association rules between the factors causing each accident and the types of high falls were obtained. The results showed that insufficient safety technology training and untimely elimination of hidden danger in safety production are the most frequent causes of accidents in fall accident reports. There are complex and close interactions between the different levels of causes in the accident report. Through association rules, the combined cause factors of 6 types of falling accidents are revealed, which can point out the direction for the development of falling accident emergency plans and accident investigation under different circumstances. This study can scientifically and reasonably dig out the hidden rules of high-fall accidents, and provide a theoretical basis for preventing these accidents during construction projects.