基于边坡位移监测数据的进化支持向量机预测模型研究
Support vector machine prediction model based on slope displacement monitoring data
-
摘要: 支持向量机方法是基于统计学习理论和结构风险最小化原则的学习方法,在回归预测方面具有良好外推能力,并且适合小样本的统计学习问题。建立支持向量机预测模型,对边坡位移进行预测计算,将预测值和实测值对比分析,验证了支持向量机预测模型较强的外推能力和预测计算的有效性。通过对边坡位移初始时序位移数据进行灰色理论的累加生成和累减生成处理,形成新的时间序列数据,在此基础上,计算出预测值,并与基于初始时间序列的支持向量机预测结果对比分析,基于新生成的时间序列数据进行预测计算结果精度明显提高。基于边坡位移监测数据构建训练样本数据集,研究了训练样本数据集的选取对预测结果的影响。对支持向量机预测模型的关键参数进行敏感度分析,并采用进化算法–微粒群算法对支持向量机模型参数加以优化,提高了预测精度。Abstract: Based on the statistical learning theory and the principle of the minimum structural risk,the support vector machine(SVM) method has the excellent extrapolating ability for regression prediction and good applicability to the problem on small training data.The extrapolating ability and predicting capability have been validated by comparing the monitoring values and predicted ones obtained by using the support vector machine prediction model built on displacement monitoring data of the slope project.Based on the new data obtained by generating operation on the initial monitoring data of the slope project,the predicted results are figured out with SVM model correspondingly and good prediction precision is approved as well by comparing the predicted results based on the initial data and new data.The influence of the selected training data on the prediction precision is also analyzed.The sensitivity analysis of the parameters of SVM model is made as well.Moreover,the precision of prediction is improved by using one of evolutionary algorithms,particle swarm optimization algorithms,to optimize the key parameters of SVM model.