To address the limitations of conventional empirical formula-based pore pressure prediction methods in engineering practice, such as high dependency on velocity, numerous required empirical parameters, and significant human influence, this study proposes an intelligent pore pressure prediction model based on eXtreme Gradient Boosting (XGBoost). By incorporating the ratio of actual P-wave velocity to the normal compaction trendline as a feature parameter in model training, the prediction accuracy and generalization capability of pore pressure are significantly improved. Furthermore, an enhanced method is introduced, which replaces the normal compaction trendline with the Dv curve for pore pressure prediction, effectively mitigating the computational complexity and subjectivity associated with establishing the normal compaction trendline. The effectiveness of this improved method is also validated across other machine learning regression models. The results demonstrate that the proposed intelligent pore pressure prediction model and its enhanced method exhibit high prediction accuracy and generalization ability, providing efficient and reliable data support for drilling safety. This approach holds significant engineering application value and broad prospects for future use.