基于偏最小二乘回归的军用飞机采购价格预测

考虑到军用飞机采购价格样本数据少、难于预测的特点和偏最小二乘回归方法在处理小样本多元数据方面的优势,提出一种基于偏最小二乘回归的军用飞机价格预测方法.偏最小二乘回归首先提取第一、第二主成分对采购价格样本的特异点进行剔除;然后进行变量投影重要度分析来筛选变量;最后,偏最小二乘回归对筛选的变量进行回归建立军用飞机价格预测模型,并对军用飞机价格进行预测.结果表明,在军用飞机价格预测方面,与未筛选变量的回归模型和逐步多元回归相比,经过变量筛选的偏最小二乘回归模型预测的精度更高,更能体现采购价格与飞机性能参数之间的关系.

Battleplan acquisition cost forecast based on partial least-squares regression

Considering few observations and forecast difficulties of battleplan acquisition cost and the advantages of partial least-squares regression (PLSR) in analyzing multivariate data with few observations, a new method of battleplan acquisition cost forecast based on PLSR is proposed. Firstly, PLSR extracts the first and second principal components of battleplan acquisition cost data to identify the outliers in observations; secondly, under the control of cross validation, PLSR extracts principal components from valid observations and selects variables by variable importance in projection; in the end, PLSR regresses to selective variables and builds the battleplan acquisition cost forecast model. The acquisition cost of a battleplan is forecasted with this model. The results calculated show that compared with PLSR model of nonselective variables and stepwise regression model, the PLSR model of selective variables has higher accuracy in battleplan acquisition cost forecast, and can reflect better the correlation between acquisition cost and performance parameters of battleplan.