Abstract: Selecting the appropriate assessment model is crucial to the susceptibility assessment of rock collapse hazards. Taking the collapse hazards in the Longjuba area of the Modaoxi basin as an example, this article establishes a logistic regression model considering dummy variables and a logistic regression model not considering dummy variables, and the models are compared to accurately establish the susceptibility zoning. The analysis results show that according to the Cox & Snell R2 statistical results, the Nagelkerke R2 statistical results, the confusion matrix results, and the hosmer-Lemeshow results, the regression model considering dummy variables provides a better fit to the samples and a higher right percentage of classification whether collapse has occurred. In the regression model considering dummy variables, the normalized probability values of the high easy-happening area and the easy-happening area are 93%. Comparing to only 65% in the model not considering dummy variables, the former has a better zoning effect.