飞机研制费用的组合预测方法

针对飞机研制费用预测样本数据较少、影响因素复杂，单一预测方法预测质量不高的问题，采用组合预测方法预测飞机研制费用。组合了径向基函数（radial basis function, RBF）神经网络、格拉姆施密特回归、偏最小二乘回归(partial least squares regression, PLSR)3种预测方法，基于样本数据建立了飞机研制费用组合预测模型，并与单项预测进行了对比分析。结果表明,组合预测具有满意和稳定的预测精度，并可以降低单项预测的质量风险，是飞机研制费用预测可靠而有效的方法。

Combination forecasting method for development cost of aircraft

To address the problems of scarce sample data, complex influence factors and low forecasting quality of a single prediction method in predicting the aircraft development costs, a combination forecasting method is adopted. Based on the sample data, the radial basis function (RBF) artificial neural network, Gram Schmidt regression and partial least squares regression (PLSR) are combined to construct the combination forecasting model, which is also compared with the single prediction method. The results show that the combination forecasting method has satisfactory and stable prediction accuracy, and it can reduce the quality risk of the single prediction method, so it is a reliable and effective method for aircraft development costs prediction.