Abstract:

The Metro engineering projects are typically high risky and complex, of which the safety accidents occur often and the risk factor varies. 151 Metro construction safety accident reports from 2002 to 2015 are selected to identify the risk factors and provide reference for risk control. Word processing, feature selection, vector space model and cooccurrence pattern recognition are carried out for risk accident reports based on R language and text mining method; the word cloud and network structure are used to visualize text mining results. Finally, 6 key risk factors and 23 general risk factors are identified; and a risk factor set is built based on key risk factors. The results can provide reference for safety risk control of Metro construction.

Key words: Metro construction, safety accident, text mining, R language, safety risk factor