基于机器学习的工业建筑工程造价预测研究  

Forecasting the Costs of Industrial Construction Based on Algorithm Optimized Neural Network and Support Vector Machine

作　　者：陈悦华[1] 郑思敏 刘文路 CHEN Yuehua;ZHENG Simin;LIU Wenlu(School of Civil Engineering,Wuhan University,Wuhan 430072,China;不详)

机构地区：[1]武汉大学土木建筑工程学院,湖北武汉430072 [2]中信建筑设计研究总院有限公司,湖北武汉430014

出　　处：《武汉理工大学学报（信息与管理工程版）》2021年第4期314-321,共8页Journal of Wuhan University of Technology：Information & Management Engineering

基　　金：国家自然科学基金项目(71073117);武汉市城乡建设局科技项目(201929).

摘　　要：为在项目前期进行准确快速的工业建筑工程造价预测,在现有研究的基础上,对比多种算法优化的神经网络与支持向量机模型预测效果,以得到最优工业建筑造价预测模型。通过箱型图对样本进行筛选,分别利用主成分分析法(PCA)与灰色关联法(GRA)对输入指标进行处理,将处理后的数据分别导入BP、GA-BP、PSO-BP、CV-SVM、GA-SVM、PSO-SVM 6个模型进行训练与预测,对预测结果进行多方面对比分析,选择最优造价预测模型。结果显示,PCA-CV-SVM模型为最优模型,运行时间短、预测结果稳定且精确度高。In order to accurately and quickly predict the cost of industrial construction projects in the early stage of the project,based on the analysis of existing research,this paper compares the prediction effects of neural network and support vector machine optimized by various algorithms to obtain the optimal industrial construction cost prediction model truly and reliably.The samples are screened through box plots,and the input indicators are processed by principal component analysis(PCA)and gray correlation method(GRA).Then,the processed data was imported into 6 different models,including BP,GA-BP,PSO-BP,CV-SVM,GA-SVM and PSO-SVM,and the prediction results are compared and analyzed in many aspects,and the optimal cost forecast model is selected.The results show that PCA-CV-SVM model is the optimal model,which has the advantages of the shortest running time,stability and high accuracy.

关 键 词：工业建筑 造价预测 神经网络 支持向量机 优化算法 

分 类 号：TU-17[建筑科学]