基于ANN贡献分析及GEP算法的地铁车站土建造价预测模型  

Prediction model of subway station civil engineering cost based on ANN contribution analysis and GEP algorithm

作　　者：王杰[1] 卢毅[1] WANG Jie;LU Yi(School of Traffic and Transportation Engineering,Changsha University of Science and Technology,Changsha 410114,China)

机构地区：[1]长沙理工大学交通运输工程学院,湖南长沙410114

出　　处：《铁道科学与工程学报》2020年第8期2152-2161,共10页Journal of Railway Science and Engineering

基　　金：2018年度湖南省交通运输厅科技进步与创新计划项目资助(201836)。

摘　　要：科学地选取工程特征因素及预测方法对于构建一个好的造价预测模型十分关键。在选取的14个影响地铁车站土建造价特征因素中,利用神经网络贡献分析的变量选取方法,筛选出12个主要特征因素。并针对这选定的14个全部特征因素和12个主要特征因素分别组合BP神经网络和GEP 2种预测方法构建4个不同的造价预测模型,应用18组地铁车站土建造价和特征因素的历史数据进行实例探究,通过R2,MSE,RMSE和MaxRE 4个指标的评价,结果表明:用主要特征因素为模型输入变量能显著提高模型的预测精度,且和GEP算法组合建立的造价预测模型为最优。将主要特征因素选取和预测方法选取相结合构建求解的最优模型很好地解决了已有相关研究中选取特征因素主观性多科学性不足及未考虑特征因素选取对预测方法选取的影响问题。Selecting engineering characteristic factors and forecasting methods scientifically is very important for constructing a good cost forecasting model.Among the 14 characteristic factors influencing the construction cost of subway stations,12 main characteristic factors were selected by using the variable selection method of neural network contribution analysis.According to allthe14 characteristics of the selected factors and 12 key features combined BP neural network and GEP respectively,two types of forecast method were used,and four different cost prediction models were built.The18 sets of subway station construction cost factors and characteristics of the historical data were used to perform example exploration,and to conduct evaluation through R2,MSE,RMSE,MaxRE four indices.The results show that the use of the main characteristic factors as model input variables can significantly improve the prediction precision,and the cost prediction model with the combination of GEP algorithm is the most optimal.By combining the selection of main characteristic factors with the selection of prediction methods,an optimal model was constructed to solve the problem of lack of subjectivity and multi-scientificity in the selection of characteristic factors and the influence of the selection of feature factors on the selection of prediction methods.

关 键 词：交通运输经济 特征因素 造价预测 GEP模型 ANN模型 地铁车站 

分 类 号：U231.4[交通运输工程—道路与铁道工程]