Combining with the characteristics of aviation ammunition training consumption, this paper studies the forecasting of aviation ammunition training consumption based on the fusion of neighborhood rough sets (NRS) and support vector machines (SVM). The five initial influencing factors are reduced to three core influencing factors by using NRS and the training set is used to optimize SVM. The best penalty parameter and kernel parameter are obtained through parameter optimization and then the NRS-SVM combination forecasting model is constructed to forecast the training consumption of aviation ammunition. Empirical studies show that predictions using the model are in good agreement with the actual data and the model has better predictive performance than the other prediction models.