一种基于Adaboost和变量筛选的LSSVM工程造价估计方法  

An Estimation Method of Engineering Cost Based on Adaboost and Variable Selection with LSSVM

作　　者：黄文涛[1] 周萍 程锦翔[3] 

机构地区：[1]东南大学电气工程学院,江苏南京210096 [2]建业恒安工程管理股份有限公司,江苏江阴214400 [3]南京航空航天大学机电学院,江苏南京210016

出　　处：《重庆交通大学学报（自然科学版）》2016年第3期54-57,104,共5页Journal of Chongqing Jiaotong University(Natural Science)

基　　金：江苏省青年自然科学基金(BK20140538)

摘　　要：为了实现利用较少的工程信息,快速准确的估计出工程项目的造价,提出了一种基于Adaboost-VIP的工程造价估计方法。首先采用变量投影重要性指标(variable importance in projection,VIP)法对影响工程造价的多个因素进行特征提取,然后利用最小二乘支持向量机作为非线性逼近器,建立工程造价的估计模型。为了进一步提高模型的估计精度,将自适应提升算法(Adaboost)与VIP相结合,利用Adaboost将多个弱造价估计模型进行集成,得到强造价估计模型。同时将该方法应用到建筑案例中,结果表明:VIP方法能有效地对影响因素进行筛选,简化模型结构;Adaboost-VIP模型与单一的工程造价估计模型相比,具有更高的估计性能。To realize the fast and accurate prediction of construction engineering cost by using less engineering information, a novel estimation method of engineering cost based on Adaboost-VIP was proposed. Firstly, variable importance in projection （VIP） method was used to extract the multiple factors affecting the engineering cost, and then least squares support vector machine （LSSYM） was used as a nonlinear approximation to establish the estimation model of engineering cost. In order to further improve the estimation precision of the model, Adaboost method was combined with VIP. Some weak predictors were integrated by Adaboost and then a strong predictor was obtained. Meanwhile, the method was applied to the case study of construction. The results indicate that： variable importance in projection method can effectively choose the key influence factors and simplify the structure of the model; compared with the single engineering cost estimation model, the Adaboost- VIP model has higher estimation performance.

关 键 词：道路工程 工程造价 自适应提升法 变量投影重要性指标 最小二乘支持向量机 

分 类 号：U415[交通运输工程—道路与铁道工程] TU723[建筑科学—建筑技术科学]