Accurate measurement of geometric errors and mass properties is essential to aero-engine manufacturing. The measurement results can provide a reference for assembly. To address the problem of poor parallelism and unbalanced prediction accuracy of hyperbolic paraboloid rotors, this article designs a classifier. It achieves precise differentiation of the rotor surface by combining the principal component analysis (PCA) and support vector machine (SVM). The results show that the surface classifier accuracy is 99%. Then this article combines the genetic algorithm (GA) and back propagation neural network (BPNN) to predict the parallelism and unbalance of the hyperbolic parabolic rotors. Compared with the traditional prediction model, the GA-BP neural network is more accurate in predicting the parallelism and unbalance of the multistage assembled rotors. The trend of the predicted results is more consistent with the experimental data. The prediction error of parallelism is reduced by $7.5~\mu \text{m}$ and the maximum deviation of unbalance is reduced by 250.6 $\text{g}\cdot$ mm through the GA back-propagation (GA-BP) neural network. The surface classifier and intelligent algorithm designed in this article can provide a reference for rotor manufacturing and assembly.