The leaves of tillering onions are tubular, and conventional photosynthesis measurement instruments are unable to accurately gauge the net photosynthetic rate using their leaf chamber clamps. To address this issue, this paper utilizes hyperspectral reflectance combined with machine learning algorithms to quantitatively analyze the net photosynthetic rate in the leaves of tillering onions. Principal Component Analysis（PCA） was employed in conjunction with Random Forest（RF） and Support Vector Machine（SVM） models to effectively extract spectral features indicative of the net photosynthetic rate from hyperspectral data, establishing a model for the inversion of net photosynthetic rate based on hyperspectral reflectance. Principal Component Analysis effectively reduced the dimensionality of the data, capturing 92.9% of the variance within the first nine principal components, thereby providing a solid foundation for the machine learning models. In the tested models, the Random Forest model performed excellently,with a coefficient of determination（R²） of 0.94 and a root mean square error（RMSE） of 1.5154, demonstrating high prediction accuracy and stability. Although the Support Vector Machine model had a lower R² of 0.81 and an RMSE of 1.63, it was still proven capable of handling complex nonlinear relationships in the data, showing robustness and generalization ability in predictive modeling tasks. Comparative analysis indicates that the Random Forest model is more suited for constructing models for the inversion of net photosynthetic rate using hyperspectral reflectance. This model reduces predictive errors through an ensemble strategy, providing more stable and accurate estimates. Therefore, the methodology employed in this study represents an effective approach for estimating the net photosynthetic rate in the tubular leaves of tillering onions.