Abstract: To elucidate the formation mechanism of the closed pressure drop of cigarette rods and to develop a predictive model, CT-based three-dimensional reconstruction technology, physical model, and gradient boosting decision tree algorithm were jointly employed to systematically analyze the closed pressure drop of 34 cigarette rod samples. Data quality was improved through preprocessing using a physical model, and the predictive performances of four types of models, namely traditional physical model, basic theoretical model, empirical-physical model, and machine learning model, were systematically compared. Using cigarette rods with a circumference of 20 mm as a case study. The results showed that: 1) Among the three types of physical models, the empirical-physical model showed the best predictive performance and provided a basis for data cleaning and feature screening. Using this model, outliers with residuals exceeding 100 Pa, a threshold of approximately 2.6 times the median absolute deviation, were identified and removed, thereby improving the physical plausibility of the dataset. 2) The gradient-boosted decision tree model achieved the best predictive performance on the independent test set (MAE=38.58 Pa, R² =0.850, MAPE=7.08%), outperforming other types of physical models. 3) The equivalent pore diameter extracted from three-dimensional reconstruction was the dominant factor affecting the closed pressure drop, with a feature importance of 76.73%. Removing this feature increased the mean absolute error of prediction by 78.6%. 4) The equivalent aperture was significantly influenced by the interaction between the width of cut tobacco and stalk position of tobacco leaves, suggesting that the closed pressure drop can be targetedly controlled through process adjustments. Although validated only for demi-slim cigarettes, the proposed physically guided and data-driven integrated framework provides a new method for optimizing the entire process from raw material to processing in the tobacco industry, and offers a reference for the flow analysis of porous media.