基于优化支持向量机及极限学习机的滑坡变形趋势研究

Landslide Deformation Trend Based on Optimal Support Vector Machine and Extreme Learning Machine

受地质环境及多种诱发因素的影响,滑坡变形包含了多层次的信息,通过单一预测模型难以准确、有效地判断滑坡变形趋势.因此,建立多方法集合的系统模型十分必要.首先,利用小波去噪对滑坡变形序列进行去噪处理,将滑坡变形分解为趋势项和误差项,采用PSO-SVM模型对滑坡变形的趋势项进行预测,利用ELM神经网络进行误差修正,两者结果叠加即得到滑坡的变形预测值;其次,利用秩相关系数检验对滑坡的变形趋势进行判断,以验证前法预测结果的准确性,并探讨该方法在滑坡变形趋势判断中的适用性.经实例检验,得出:该预测模型的预测值与实测值基本相符,且较单一预测模型具有更高的预测精度;同时,秩相关系数检验的结果与预测结果相符,验证了变形预测结果的准确性及该方法在滑坡变形趋势判断中的适用性.

Affected by the geological environment and a variety of factors,landslide deformation includes various levels of information so that through a single landslide forecasting model it is difficult to accurately and effectively judge the landslide deformation trend. Therefore, the system model set is very necessary. First of all, the landslide deformation is divided into trend item and the error term by using wavelet denoising on the landslide deformation sequence denoising, and by using PSO-SVM model to predict the landslide deformation trend,the error is corrected by the ELM neural network,and the both are added to obtain the value of landslide deformation prediction. Secondly,we use rank correlation coefficient to test the deformation trend of the landslide and to determine the accuracy of the above prediction result,and to explore the applicability of the landslide deformation trend of the judgment. The practical example shows that the prediction values from the model are consistent with the measured values, and compared with the single prediction model it has higher prediction accuracy. At the same time,the rank correlation coefficient test and prediction results are consistent to verify the accuracy of the results and the applicability of the method in landslide deformation trend judgment.