【摘要】 对于城镇自来水厂建设项目,项目前期准确的造价估算是项目能够发挥预期效益的保证。由于项目前期缺少详细的设计文件,同时涉及多种复杂且不确定的因素,且造价与各因素间的关系并不固定,在进行水厂项目造价估算时,大多需要依赖历史项目样本数据的建模方法。文中主要基于实际自来水厂建设项目工程造价数据,分析筛选出18个与工程造价相关的影响因素作为输入变量,分别借助人工神经网络中的BP和RBF算法,通过对人工神经网络模型的校验和对比,发现2种模型对训练样本数据都具有很好的拟合性。研究表明,BP神经网络模型具有更好的预测能力,所有测试样本的估算精度可以控制在±30%以内,达到了项目建议书阶段的估算精度要求。更多还原

【Abstract】 For construction project of WTP, accurate initial cost estimates will be a guarantee of the project profitability. Due to the lack of detailed design documents, with multiple complex and uncertain factors, there is no theoretical formula between construction cost and each factor during the early stages of the project. Based on actual construction cost data from WTP projects, 18 influencing factors related to construction cost are selected as input variables. Using back propagation(BP) and radial base function(RBF) arithmetic in artificial neural network(ANN), through verification and comparison of the two kind of ANN models, it is found that both models have a good fit to the training sample data. Studies show that BP ANN model has better prediction accuracy, and estimation accuracy of all test samples can be controlled within ±30%, which meets estimation accuracy requirements of project proposal stage.更多还原