摘要
影响木结构古建筑寿命的因素主要包括物理、化学和生物等多个方面,且影响过程是错综复杂的。这些因素之间可能非线性相关,相应观测数据特征提取的完整性决定了其寿命预测的精确性。文中引入基于样条变换的偏最小二乘预测方法,在更高维空间中把非线性预测模型的建立问题转化为线性预测模型的建立问题实现木结构古建筑寿命预测。针对观测数据缺失情况下,基于样条变换的非线性偏最小二乘方法预测精度较低的问题,文中给出一种基于缺失数据估计的非线性偏最小二乘预测方法,以更充分地抽取可用观测数据的特征信息,并将其应用于木结构古建筑寿命预测中。仿真和实验验证结果表明了该方法的有效性。
The factors that affect the life of the ancient building of the wood structure include many aspects, such as physical, chemical and biological factors, and the process of impact is complicated. These factors may be nonlinear correlation, and the completeness of the corre- sponding observation data feature extraction determines the accuracy of the life prediction. In this paper,the PLS based on spline transfor- marion is introduced, and the problem of building the nonlinear model in higher dimensional space is transformed into it of building the linear model to realize the life prediction of wood structure. In order to estimate the accuracy of the nonlinear PLS method based on the case of the missing data, a nonlinear PLS prediction method is given based on the missing data estimation to extract the feature informa- tion of the available observation data more adequately and used in the life prediction of wood structures. Simulation and experimental veri- fication results show the effectiveness of the proposed method.
出处
《计算机技术与发展》
2016年第6期195-199,共5页
Computer Technology and Development
基金
国家自然科学基金资助项目(61203094
61174112)
河南省高校科技创新人才支持计划(2012HASTIT005)
关键词
数据缺失
样条变换
偏最小二乘回归
古建筑
寿命预测
data missing
spline transformation
Partial Least-Squares Regression （PLSS）
ancient building
life prediction
