The Earth Pressure Balance (EPB) shield machine plays a pivotal role in underground tunnel excavation, where precise control of chamber pressure is essential for maintaining tunnel stability and minimizing risks. Traditional EPB control methods heavily rely on operator experience, resulting in delays and limited responsiveness to sudden geological changes. This paper presents an improved EPB mechanism model that builds upon traditional approaches, which primarily consider chamber pressure changes caused by soil volume variations. The improved model further incorporates the effects of excavation face pressure variations, arising from factors such as cutterhead soil extrusion and changing geological conditions. By integrating these additional influences, the model achieves more accurate predictions of chamber pressure. To further enhance performance, a hybrid modeling approach is proposed, combining the improved mechanism model with a data-driven component that compensates for residual prediction errors. The hybrid model is validated using field data from two distinct tunneling projects, demonstrating superior prediction accuracy and generalization capability compared to standalone mechanisms and data-driven models. The results confirm that the proposed hybrid model significantly improves pressure prediction accuracy and provides a more reliable solution for intelligent control of the EPB process.