Due to the complex and dynamic nature of construction environments, safety accidents occurring in these environments pose a grave threat to life and property. Therefore, it is essential for safety managers in construction, supervisory, and related units to adopt a rigorous and systematic methodology for assessing the risks associated with construction safety accidents. This will enable managers to comprehend the likelihood of accidents, subsequently enabling them to implement preemptive and control measures to reduce the probability of such incidents. Drawing on the accident causation theory, this study utilized web crawler technology to collect construction accident reports, subsequently employing text mining (TM) techniques to identify the accident causal factors specified in 166 accident reports. Subsequently, 33 key features were extracted from the accident causal factors, and correlation rule mining was used to analyze the correlations between the causal factors. Successively, a Bayesian network (BN)-based risk assessment model was constructed for construction safety accidents. Finally, through reverse reasoning, this study identified the probable paths of construction safety accidents and the sensitive factors that trigger such accidents. The results showed that management factors (MFs) are the primary drivers of accidents, highlighting the importance of focusing on preventive and control countermeasures for factors characterized with high severity and sensitivity.