Abstract: Unsafe human behavior is an important cause of safety accidents. In order to study the factors that affect the unsafe behavior of water conservancy construction personnel, this article analyzes 280 water conservancy construction accident cases using text mining technology, and uses R language to form a visual word cloud graph to obtain 32 factors that affect unsafe behavior. Using Ucinet software to construct co-occurrence networks and calculate centrality, identify key factors affecting unsafe behavior of water conservancy construction personnel.