| 进化多核相关向量机的大坝变形预测 |
| |
| 引用本文： | 张立亭,徐志宽,罗亦泳,况达,李剑.进化多核相关向量机的大坝变形预测[J].测绘科学,2017,42(7). |
| |
| 作者姓名： | 张立亭 徐志宽 罗亦泳 况达 李剑 |
| |
| 作者单位： | 1. 东华理工大学测绘工程学院,南昌,330013;2. 武汉大学测绘学院,武汉,430079 |
| |
| 基金项目： | 国家自然科学基金项目，江西省自然科学基金项目，江西省教育厅科学技术研究重点项目 |
| |
| 摘 要： | 针对大坝变形系统的非线性、复杂性以及不确定等特点,提出一种优化多核相关向量机的大坝变形预测模型方法。通过对实验数据进行归一化处理,核函数的加权组合以及遗传算法对模型参数的优化,建立遗传算法优化多核相关向量机的大坝变形预测模型。实验结果表明:数据归一化能归纳统一样本的统计分布性,加快梯度下降求解最优解速度和提高预测精度;优化的加权核函数能有效提高模型预测精度;各项精度指标值均优于BP神经网络方法、多项式核相关向量机方法预测精度,证实优化的多核相关向量机模型是一种精度较高的大坝变形预测方法。
|
| 关 键 词： | 大坝变形 遗传算法 多核函数 相关向量机 |
Dam deformation prediction based on evolutionary multiple kernels relevance vector machine |
| |
| ZHANG Liting,XU Zhikuan,LUO Yiyong,KUANG Da,LI Jian.Dam deformation prediction based on evolutionary multiple kernels relevance vector machine[J].Science of Surveying and Mapping,2017,42(7). |
| |
| Authors: | ZHANG Liting XU Zhikuan LUO Yiyong KUANG Da LI Jian |
| |
| Abstract: | In terms of the characteristics of nonlinear,complexity and uncertainty of the dam deformation,a method for predicting dam deformation that based on optimization of multi-kernel relevance vector machine is presented.By the normalizing experimental data,weighted combination of kernel functions and genetic algorithm optimization model parameters,finally the model of dam deformation prediction was established,which based on relevance vector machine of multi-kernel of genetic algorithm optimized.The experimental data results show that the data normalization that can be summarized unified statistical distribution of the sample,accelerating gradient descent achieve solve the optimal speed;Optimization of weighted kernel function which can improve the accuracy of model prediction effectively;The precision index value are all better than the prediction accuracy of PSOBP method and single-kernel RVM method,it is proved that the optimized multi-kernel RVM model which is a higher accuracy method to predict the dam deformation. |
| |
| Keywords: | dam deformation genetic algorithm multi-kernel function RVM |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
|
