摘要
针对变形序列呈现非线性、随机性等特点,提出一种基于小波分析的灰色支持向量机变形预测新算法。首先采用小波原理对变形序列进行分解,得到不同频率的分量,根据各分量的特征分别建立灰色模型和支持向量机模型。同时,采用网格搜索法选取模型的最优参数,并与灰色模型、BP神经网络和支持向量机对比分析。经理论分析和算例表明:采用小波原理能够有效提取变形序列中隐含的时频信息,优化了原始序列,能更好地反映变形的局部特征和变化趋势;新算法继承了支持向量机泛化能力强、非线性拟合好等优良特性,避免了灰色模型存在的缺陷,保证了较优的局部预测值和较好的全局预测精度,应用变形预测是可行的。
In view of the non-linearity and randomness of deformation sequences,a new algorithm for deformation prediction of grey support vector machine based on wavelet analysis is proposed. Starting from the time-frequency analysis,the wavelet principle is used to preprocess the deformation sequence and obtain the components with different frequencies. The gray model and the support vector machine model are established according to the characteristics of each component. At the same time,the grid search method is used to select the optimal parameters of the model,and compared with the gray model,BP neural network and support vector machine. Theoretical analysis and examples show that the wavelet principle can effectively extract the implicit time-frequency information in the deformation sequence,optimize the original sequence,and can better reflect the local characteristics and the changing trend of the deformation; the new algorithm inherits the support vector machine. It has excellent features such as strong ability of adaptability and non-linear fitting,avoids the defects of the gray model,guarantees better local prediction value and better global prediction accuracy,and it is feasible to apply deformation prediction.
作者
陆杰
覃书林
徐宁辉
Lu Jie, Qin Shulin, Xu Ninghui(Nanning Stow-eying and Mapping Geographic Information Institute ,Nanning 530001, China)
出处
《城市勘测》
2018年第5期153-157,共5页
Urban Geotechnical Investigation & Surveying
基金
广西空间信息与测绘重点实验室项目(16-380-25-22)
关键词
变形监测
小波分析
灰色模型
支持向量机
精度评定
deformation monitoring
wavelet analysis
grey model
support vector machine
accuracy evaluation
