Causative word extraction and analysis based on text data of metro safety events

China’s intercity and urban rail transit has entered a stage of unified construction and rapid development, and its safety control is difficult to coordinate, and all kinds of safety events or failures always occur during the metro operation. When facing a large amount of text data such as subway accident reports, it is necessary to extract information from them in order to get the most critical and useful information quickly. In order to help the relevant personnel to efficiently find the causes of accidents and put forward improvement measures, this paper takes the unstructured text data of subway as the research object. At the same time, it considers the word network and sentence network, constructs a text two-layer network to study the importance of the text, and puts forward the relevant text feature extraction method and the causal text recognition method to find out the key factors of accident causation. Finally, the method is applied to analyze the causal factors of the safety accidents in Wuhan metro in the past 10 years, to find out the maximum possible causal chain and the biggest safety risks for the metro staff, and to remind them to enhance their management ability to prevent accidents effectively.