论文部分内容阅读
支持向量机(Support Vector Machine,SVM)已被广泛应用到滑坡位移预测,但在具体使用时,SVM的惩罚系数C、核函数参数δ及松弛系数ζ这三个重要参数的取值选择成为影响预测精度的关键。为有效分析SVM三参数取值对滑坡位移预测精度的影响规律,以三峡库区浮托减重和动水压力型两类典型水库滑坡为代表的连续6年地表位移、降雨及库水位监测数据为研究对象,首先,采用移动平均法将位移数据分解为趋势项和周期波动项,并区分训练集和检验集;再结合对滑坡变形机理及影响因素的分析,选择相应预测变量分别建立趋势项和波动项位移预测SVM模型;然后,在固定两参数情形下,通过改变另一参数的取值大小以获得SVM训练集与检验集的预测精度变化规律;最后,建立起典型水库滑坡SVM位移分解预测的参数取值推荐范围。该取值范围可以作为滑坡位移预测SVM模型的参数寻优初始搜索范围,可以在保证预测精度的前提下大大提高搜索效率。
Support Vector Machine (SVM) has been widely applied to landslide displacement prediction. However, in the specific use, SVM penalty coefficient C, kernel function parameter δ, and relaxation factor ζ are three important parameters of choice The key to forecast accuracy. In order to effectively analyze the influence of three parameters of SVM on prediction accuracy of landslide displacement, six consecutive years of surface displacement, rainfall and reservoir water level monitoring data, represented by two types of typical reservoir landslides with floating weight loss and hydrodynamic pressure in the Three Gorges area, Firstly, moving data are used to decompose displacement data into trend items and periodical fluctuation items, and to distinguish training set and test set. Combined with the analysis of landslide deformation mechanism and influencing factors, the corresponding predictive variables are selected to establish trend items And the fluctuant term displacement prediction SVM model. Then, by changing the value of the other parameter under the fixed two-parameter condition, the variation rule of the prediction accuracy of SVM training set and test set is obtained. Finally, the typical SVM displacement decomposition of reservoir landslide The predicted parameters are recommended range. The range of values can be used as the initial search range of the parameters of the landslide displacement prediction SVM model, which can greatly improve the search efficiency under the premise of ensuring the prediction accuracy.