Drilling engineering is a necessary means to explore and develop oil and gas resources, overflow is a common risk in the drilling process, and once it happens, the consequences will be extremely serious underground complex accidents, if the discovery is not timely, improper treatment methods will even cause blowout, well collapse and other serious accidents. Therefore, it is very important to build an efficient overflow prediction model and prevent well control accidents in early detection period. The purpose of this paper is to establish a random forest model to identify drilling overflow risk. First, a large number of drilling parameter data were collected to build a data set for overflow risk identification. Secondly, data pre-processing and feature engineering are carried out. The pre-processing mainly includes data cleaning, missing value processing and noise processing. At the same time, Pearson correlation coefficient is used to select the most suitable feature value of overflow risk. Finally, the framework parameters, decision tree parameters and the most basic four parameters are set by using the classical function of random forest, and the attributes of each parameter are determined to establish the random forest model. Because random forest model is relatively simple and not easy to overfit, and has good tolerance in terms of prediction accuracy, anti-outlier points and anti-noise, it can be effectively applied to identify the overflow risk in drilling operations and reduce the occurrence of drilling accidents and economic losses.