基于Stacking融合的工程造价预测组合模型

Combination Model of Project Cost Prediction Based on Stacking Fusion

为了快速准确地进行工程造价预测,本文收集高层住宅工程造价历史数据,利用相关性系数法对定量指标进行约简,运用Python构建了以随机森林、XGBoost和岭回归作为初级学习器,岭回归作为元学习器的Stacking集成学习的融合模型。结果表明:基于Stacking融合模型预测精度较高,结果稳定,平均绝对误差在5%以内,有助于项目建设前期的造价预测。

In order to predict the construction project cost quickly and accurately,the historical data of high-rise residential project cost is collected,and the quantitative indexes are reduced by using the correlation coefficient method.Python is used to construct a fusion model of Stacking integration learning with random forest,XGBoost and ridge regression as primary learners and the ridge regression as a meta-learner.The results show that the fusion model based on Stacking has high prediction accuracy and stable results,and the average absolute error is controlled within 5%,which is helpful to the cost control in the early stage of project construction.