小样本、贫信息下民用飞机费用估算模型及算法

Civilian Aircraft Cost Estimation Model and Algorithm Based on Small Sample and Poor Information

费用估算是民用飞机成本管理的核心内容,直接关系到民用飞机项目的成败。针对民用飞机费用估算问题展开研究,把GM(0,N)模型与BP神经网络算法相结合,应用BP神经网络对建立的GM(O,N)模型估计值进行再次优化,二者的组合模型能有效提高估算精度。模型构建步骤如下:首先构建原始GM(0,N)模型并根据该模型求出样本的费用估算值;以GM(0,N)模型的参数变量、样本估计值为输入,实际值为期望输出建立BP神经网络并在mat lab里训练;利用训练后的组合模型对民用飞机的费用进行估算。以民用飞机费用估算为仿真实例,分别使用多元线性回归、传统的GM(O,N)模型,和该方法建立费用估算模型,对目标机种的费用做出估算,分析估算结果,表明该模型具有更好的拟合与估算效果。

Cost estimation is the core content of civil aircraft cost management, is directly related to the success or failure of the project of civil aircraft. The problem of civil aircraft cost estimation was studied, GM （0, N） model and BP neural network algorithm were combined, and then BP neural network algorithm was used to optimize the estimate value of GM （0, N）. The combination of the two models could effectively improve the estimation accuracy. Model construction steps are as follows： GM （0, N） model was firstly constructed and the estimate value was obtained; then, with the correlated parameters of GM （0, N）, model valuations as input, the actual value as expected output, a BP neural network was constructed. After that, using BP neural network constructed by Mat lab simulation training and the trained combination model, the cost of civilian aircraft was predicted. Taking simulation of civilian aircraft cost estimation for example, cost estimation model was constructed with multiple linear regression, traditional GM （O, N） model and the model proposed, respectively. Then the cost of target aircraft was predicted. The simulation results show that the model has better performance of fitting and prediction.