基于机器学习算法的综合管廊土建工程费预测研究  

Prediction of Civil Engineering Costs of Utility Tunnels Based on Machine Learning Algorithm

作　　者：王雪 汤志立 徐千军[3] Wang Xue;Tang Zhili;Xu Qianjun(Beijing Municipal Road and Bridge Technology Development Co.Ltd.,Beijing 100037,China;Beijing Information Infrastructure Construction Co.Ltd.,Beijing 100068,China;Tsinghua University,Beijing 100084,China)

机构地区：[1]北京市政路桥科技发展有限公司,北京100037 [2]北京信息基础设施建设股份有限公司,北京100068 [3]清华大学,北京100084

出　　处：《工程造价管理》2025年第1期32-37,共6页Engineering Cost Management

基　　金：国家自然科学基金项目(52090084)超大城市深层地下空间韧性基础理论,2020年。

摘　　要：开展综合管廊工程投资预测对支撑综合管廊可行性论证具有重要意义。文章以在综合管廊投资中占比较大的土建工程费为研究对象,构建了包含98组综合管廊投资的数据集,建立了21个综合管廊土建工程费预测模型,获得了最优的预测模型。研究结果表明,构建的预测模型的确定系数在0.547~0.927之间,表现最好的预测模型为TPOP算法模型,其确定系数为0.927。研究成果发展并完善了综合管廊工程投资预测方法。Conducting investment forecasting for comprehensive pipe gallery projects is of great significance in supporting the feasibility study of comprehensive pipe galleries.The article takes the civil engineering cost,which accounts for a large proportion in the investment of comprehensive pipe galleries,as the research object,constructs a dataset containing 98 sets of comprehensive pipe gallery investments,establishes 21 comprehensive pipe gallery civil engineering cost prediction models,and obtains the optimal prediction model.The research results indicate that the determination coefficient of the constructed prediction model is between 0.547 and 0.927,and the best performing prediction model is the TPOP algorithm model with a determination coefficient of 0.927.The research results have developed and improved the investment prediction method for comprehensive pipe gallery engineering.

关 键 词：综合管廊 投资预测 机器学习 特征重要性 

分 类 号：F426[经济管理—产业经济] TU-9[建筑科学]