The stochastic new design tasks often appear during ship design, which brings some difficulties to the formulation of scheduling scheme of ship design tasks. Based on the Back Propagation (BP) algorithm, the Momentum and Self-Adaptive Learning Rate Back Propagation (MSBP) algorithm is introduced to forecast whether the stochastic new design tasks can be added into the formulated scheduling scheme of ship design tasks, so as to solve the Dynamic Scheduling of Ship Design Tasks (DSSDT) problem in the case of disturbance. In order to reduce the solving space and training difficulty, the attributes with great influence on the scheduling results are selected as the eigenvalues of MSBP algorithm. The MSBP algorithm model is constructed based on the extracted eigenvalues, and the model is trained with a large amount of data. The contrast test results show that the accuracy of MSBP algorithm is better than that of unimproved BP algorithm, and the schedulability of a stochastic new design task is most closely related to its priority.