针对军用无人机研制费用的影响因素多且复杂,样本数据又相当匮乏的问题,充分利用偏最小二乘回归方法在处理小样本数据,变量多重相关性等方面的独特优势,给出了基于偏最小二乘回归方法的预测模型和军用无人机研制费用的算法步骤.用文献[2]中的数据进行了检验,并与逐步多元回归、反向传播(BP,Back Propagation)和径向基函数(RBF,Radial Basis Function)神经网络法进行了比较,结果表明了偏最小二乘模型的实效性. In view of the fact that the cost of development of military UAV is large and complex, and the data of the sample are quite scarce, the unique advantages of partial least squares regression in dealing with small sample data and multiple correlations of variables are given. The prediction model of partial least-squares regression and the algorithm steps of military UAV development costs were tested with the data in [2], and compared with the stepwise multiple regression, backpropagation and radial The Radial Basis Function (RBF) neural network method is compared, and the results show the effectiveness of partial least squares model.