摘要
将基于随机森林结合支持向量机(RF-SVM)算法引入混凝土抗冻性研究,首先从配合比因素选取n个混凝土抗冻性影响因素,以相对动弹性模量作为混凝土抗冻性评价指标,基于原始样本利用随机森林特征选择对影响因素进行重要性评价和变量筛选,选出最优影响因素集合,作为SVM模型的训练样本,建立降维后的RF-SVM混凝土抗冻性预测模型,输出预测结果,并将其与未进行影响因素筛选的支持向量机和人工神经网络模型结果对比分析,得出RF-SVM预测结果的均方根误差最小,拟合优度最接近1,说明RF-SVM预测结果精度最高、效果最好。
In this paper,the random forest combined with support vector machine(RF-SVM)algorithm is introduced into the research of concrete frost resistance.Firstly,n concrete frost resistance factors are selected from the mix proportion factors,and the relative dynamic elastic modulus is taken as the evaluation index of concrete frost resistance.Based on the original samples,the importance evaluation and variable selection of influencing factors are carried out by using random forest feature selection,and the optimal influence factor set is selected,as the training sample of SVM model,this paper establishes the dimension reduced RF-SVM concrete frost resistance prediction model which outputs the prediction results,and compares it with the results of support vector machine(SVM)and artificial neural network(ANN)models without influencing factors screening.The results show that the root mean square error of RF-SVM prediction results is the minimum,and the goodness of fit is the closest to 1,which indicates that the RF-SVM prediction results have the highest accuracy and efficiency.The fruit is the best.
作者
张陆山
袁福银
胡毅
李铁军
吴贤国
杨赛
ZHANG Lushan;YUAN Fuyin;HU Yi;LI Tiejun;WU Xianguo;YANG Sai(No.6 Engineering Co.,Ltd.of FHEC of CCCC,Tianjin 300451,China;CCCC Road and Bridge Construction Co,Ltd.,Beijing 100027,China;CCCC Second Harbour Engineering Co,Ld.,Wuhan,Hubei 430040,China;China Communications Construction Co,Ld,Bejing 100088,China;Schoo of Ciril Engineering&Mechanics,Huazhong University of Science and Technology,Wuhan,Hubei 430074,China)
出处
《施工技术》
CAS
2020年第17期95-99,共5页
Construction Technology
基金
国家重点研发项目(2016YFC0800208)
国家自然科学基金(51378235
71571078
51308240)。
关键词
混凝土
抗冻性
随机森林
支持向量机
评价
concrete
frost resistance
random forest
support vector machine
evaluation
