In order to reduce the production safety problems caused by liquid accumulation in low-lying areas of pipelines, the minimum pressure drop gradient model and two-fluid model were used to determine the critical liquid carrying velocity. The critical liquid carrying velocity under different influencing factors was simulated by single-factor experiments, and the interaction intensity between factors was obtained by uniform experimental design. Finally, the random forest model was used to complete nonlinear data regression and field result verification. The results show that the relationship between pressure and critical liquid carrying velocity is a power function; the temperature and pipe diameter are linearly and positively correlated with critical liquid carrying velocity, and the water content and top rade share a logarithmic relationship with critical liquid carrying velocity. In terms of a single factor, the top rade has the greatest influence on the critical liquid carrying velocity, followed by water content and system pressure, and the interaction between system pressure and top rade is the strongest in terms of the quadratic factor. In the verification of gathering and transportation pipelines of 20 gas wells, the coincidence rate of the random forest model is 90%, and the coincidence rates of models by Chen Jianlei and Wang Xu are 55% and 65% respectively, which proves that random forest model can be used to predict and judge the critical liquid carrying velocity.