基于BP神经网络的大规模建设项目投资估算方法张爽(浙江国信项目管理咨询有限公司,浙江宁波315000)【摘要】目前大规模建设项目投资估算方法的估算结果与实际结果误差较大,估算速率较低。针对此问题,文章引入BP神经网络研究了一种新的大规模建设项目投资估算方法,分析市场要求,建立指数指标矩阵,确定指标和投资之间的关联度,根据关联度计算结果,通过归一化处理得到非线性算子,在BP神经网络内部通过并行处理得到输出值和输出值阈值,通过逆向分析得到连接系数,确定误差数据节点,建立估算模型,通过多次筛选得到特征算子,完成信息提取,与阈值进行对比,判断估算结果的合理性。实验结果表明,基于BP神经网络的大规模建设项目投资估算方法采用并行计算,数据拟合度能够达到99%,与传统WSR投资估算方法相比,估算速率提高了30%以上,与BIM投资估算方法相比,估算速率提高了35%,更适合于实际应用。【关键词】BP神经网络大规模建筑建设项目投资投资估算【中图分类号】TU-9【文献标识码】A【文章编号】1008-2166(2023)04-045-06【DOI】10.19730/j.cnki.1008-2166.2023-04-045AMethodforEvaluatingtheRationalityofLarge-scaleConstructionProjectCostBasedonBPNeuralNetworkZhangShuang(ZhejiangGuoxinProjectManagementConsultingCo.Ltd.,Ningbo315000,China)【Abstract】Atpresent,theerrorbetweentheestimationresultsandtheactualresultsofthelarge-scaleconstructionprojectinvestmentestimationmethodislargeandtheestimationrateislow.Inviewoftheaboveproblems,thispaperintroducesBPneuralnetworktostudyanewlarge-scaleconstructionprojectinvestmentestimationmethod,analyzesthemarketrequirements,establishestheindexindexmatrix,determinesthecorrelationdegreebetweentheindexandtheinvestment,andcalculatestheresultsaccordingtothecorrelationdegree.Thenonlinearoperatorisobtainedthroughnormalizationprocessing,theoutputvalueandoutputvaluethresholdareobtainedthroughparallelprocessinginsidetheBPneuralnetwork,theconnectioncoefficientisobtainedthroughreverseanalysis,theerrordatanodeisdetermined,theestimationmodelisestablished,thefeatureoperatorisobtainedthroughmultiplescreening,theinformationextractioniscompleted,andthethresholdiscomparedtojudgetherationalityoftheestimationresults.T...