摘要
为了解决工程造价中关键影响因素难以选择以及模型估计精度不高的问题,提出了一种基于LSSVM-SVIP的工程造价动态软测量方法。该方法通过对样本数据的重构,获得具有动态性能的工程造价初始软测量模型。同时利用改进的变量投影性指标法对工程造价初始软测量模型中多个影响因素进行筛选,得到优化模型。并以最小二乘支持向量机作为非线性逼近器,建立工程造价的软测量模型。将该方法应用于工程造价案例中,结果表明:经SVIP方法优化后的模型结构简单,估计精度高,泛化能力强,能满足实际工程要求。
In order to solve difficulty to choose key influence factors of the project cost and accuracy of model estimation,a novel dynamic soft sensing strategy based on LSSVM-SVIP for the project cost is developed. Through reconstructing the sample data,an initial soft sensing model will be obtained,which has good dynamic performance. Meanwhile,these influence factors from the initial soft sensing model are selected by using variable importance in projection improved so as to obtain an optimization model. Then,LSSVM is considered as a nonlinear approximator to establish the soft sensing model of project cost. To verify the method,a real data set of the office buildings method is applied to test the model. Simulation results show that the model optimized by SVIP method has simpler model structure and higher accuracy,which can satisfy practical engineering requirement.
出处
《工程管理学报》
2015年第5期35-39,共5页
Journal of Engineering Management
基金
江苏省青年自然科学基金项目(BK20140538)
