摘要
基于GEP编程的信息搜索能力和BP网络建立了公路工程特征因素与工程造价的关系模型,并对模型进行了最优性预测评价,从而满足工程造价预测的实际应用需求。文章运用MATLAB仿真软件获取了影响公路工程造价的12个影响因子,并运用BP神经网络训练获得了影响公路工程造价的7个主要影响因素。针对选定影响因素基于BP网络和GEP算法给出了4个不同公路造价预测模型的评价指标和优模型解的分析。研究结果表明:基于主线里程、通道数量、路基土石方量、桥梁数量、隧道数量、地貌特征、利息率7个主要特征因素的两种算法模型所获得的预测结果精度要显著高于考虑全部特种因素的模型结果;基于主要特征因素的GEP网络算法模型获得的预测结果更为精确,模型最优;应用GEP网络在处理公路工程造价这类非线性空间全局搜索中具备了很高的搜索效率,能有效弥补BP网络泛化能力较低的问题。
This article is based on GEP programming of information search capability and BP network to establish the relationship between highway engineering characteristics and engineering cost model,and the model was the optimal prediction evaluation,thus meet the demand of practical application of engineering cost prediction. The paper USES MATLAB simulation software to obtain 12 influencing factors which affect the cost of highway construction,and USES BP neural network training to obtain the seven main influencing factors that affect highway project cost. Based on BP network and GEP algorithm,the evaluation indexes of four different highway cost prediction models and the analysis of optimal model solutions are given. Results show that based on the main line mileage number,channel number,amount of earth work,bridge,tunnel number,geomorphic features,the interest rate seven main factors of two kinds of algorithm model for the prediction precision of the model is significantly higher than to consider all the special factors results; The GEP network algorithm model based on the main characteristics is more accurate and the model is optimal. The GEP network has high search efficiency in the global search of highway engineering cost,which can effectively make up the problem of low generalization ability of BP network.
出处
《公路工程》
北大核心
2018年第1期206-210,共5页
Highway Engineering
基金
重庆市教委2015年度科学技术研究项目(KJ1504208)
作者简介
郑晓蕾（1984-）,女,山东烟台人,硕士,副教授,研究方向工程造价、交通与规划.
