基于概率统计分析的SVM模型在大坝变形预测中的应用

摘要

The impact factors of dam deformation need to be input the model in the process of forecast the dam de-formation based on the SVM. The impact factors were not analyzed,and directly be input into the SVM. These input fac-tors will cause redundancy. At the same time,the relationship between impact factors is relevant,and this will cause du-plicate effect of input. There is another thing;it is hard to keep the data integrity of input impact factors. This paper in-troduces the probabilistic principal component analysis( PPCA)to the preprocessing of impact factors. The missing data can be selected based on the PPCA,and the selected impact factors are the input factors of the SVM model. The deter-mined PPCA-SVM model can be used to fit and predict the deformation,and the results of dam deformation can be ac-quired based on this PPCA-SVM model.%支持向量机对大坝变形进行预测的过程中，需要输入影响因子，如果不对影响因子直接处理，直接将大量的影响因子输入到支持向量机模型（ SVM）中，势必会造成数据的冗余，同时，影响因子与影响因子之间存在着关联性，从而造成影响因子输入的重叠效应，另外，在输入影响因子过程中，很难保证影响因子输入的完整性，针对此，本文将概率统计的主成分分析（ PPCA）引入到大坝变形影响因子预处理中，利用该模型可实现对缺失数据的提取，并将提取的数据作为SVM模型的输入因子，再利用设定的模型进行数据的拟合及预测，便可获得较为稳定的大坝变形分析结果。