建设工程项目工序的LS-SVM工期预测模型

Forecast Model of Activity Duration Based on LS-SVM in Construction Engineering Project

鉴于传统工期预测的模糊性和随机性,分析影响工程项目工期的因素及参数的获取方式.采用最小二乘支持向量机(LS-SVM)构建建设工程项目工序工期的预测模型,并用工程实例论证方法的有效性.结果表明,对类似工程或者同一工程项目的类似工序的进度执行状况进行学习,采用LS-SVM的工期预测模型预测即将开展的工程项目的工序工期,符合实际工期控制的要求.与基于BP神经网络工期预测模型对比分析,LS-SVM的工期预测模型的预测误差更小,平均训练时间更短,网络总误差更小.

Referring to fuzziness and randomness for activity duration forecast of construction engineering project by traditional ways,the influence factors of activity duration are analyzed,and parameter calculation is also proposed.A forecast model of activity duration based on the least square support vector machine（LS-SVM） is set up,and the analysis of a subway case confirms validity of this model.The model is trained by the schedule execution situation of the activities in other similar projects or the similar activities in the same project,the activity duration simulated by the model conforms with the request of schedule controlling.In the forecast model of activity duration based on LS-SVM,the prediction and network total errors is less,training time is shorter than the ones in he forecast model based on BP.