Accurately predicting formation pressure is crucial in drilling operations. It is the data foundation for ensuring work safety, optimizing drilling parameters, improving work efficiency, and reducing non-production time and operating costs. Therefore, this article studies a formation pressure prediction method mainly based on logging data. Firstly, a data preprocessing method suitable for on-site applications has been established. Optimize the input characteristics of the formation pressure prediction model. Secondly, based on mutual information analysis, optimize the model's input parameters. Finally, a formation pressure prediction method based on GA-BP was established. At the same time, a method for identifying the mechanism of abnormal pressure formation based on random forest was established to clarify the mechanism of formation pressure. This method not only comprehensively considers the influence of various measurement parameters on formation pressure, but also considers the determination of the cause of abnormal pressure. The case analysis results show that this method accurately predicts formation pressure. Can distinguish the types of causes of abnormal pressure formation.