基于SVM?RF的泥石流窗口坝闭塞度判别研究?

为了研究野外泥石流防治工程中窗口坝的开口闭塞类别,基于量纲分析理论,以室内水槽试验模拟实际工程,分析模型试验与实际工程的相关物理量及对应的相似准数;引入支持向量机和随机森林分类模型,在开源机器学习工具Scikit-Learn中,采用python编程实现算法;以室内水槽试验数据作为支持向量机和随机森林的训练样本,进行机器学习得到分类模型,提出一种用于判别泥石流窗口坝闭塞类型的新方法;将测试结果与经验公式中闭塞度判别值F的分类结果进行正确率对比,结果表明,F值的分类准确率为88%,而支持向量机为92%,随机森林为94%,随机森林分类效果最好,机器学习理论为泥石流窗口坝在实践中的设计提供了新思路。

Research on the Occlusion of Debris Flow Window-frame Dam based on SVM and RF Methods

In order to investigate the window-frame block categories of the field debris flow prevention engineering projects, the laboratory flume experiment is conducted to simulate the actual condition, using dimensional analysis method to ensure the similarity criterion of model test and the actual engineering. This research introduces the basic theory of support vector machine and random forest and realizes the algorithm in python language environment through the open source machine learning tool Scikit-Learn. Making the laboratory flume experiment data as the training sample of support vector machines and random forests then, got the learning classification model, put forward a new method for identifying debris flow dam block type. The test results compared with the empirical formula of discriminant value F block degree of accuracy of the classification, the results show that the F value of classification accuracy is 88%, and the support vector machine (SVM) was 92%, the random forest was 94%. The random forest classification effect is best. The machine learning theory provides a new idea on the debris flow window-frame dam design in practice.