摘要
LNG接收站工程投资高、周期长,在前期阶段开展精准高效的造价预测至关重要。本文对如何利用历史项目数据进行机器学习,实现LNG接收站工程造价预测进行探讨。首先通过文献分析确定适用于小样本数据的SVM和GBDT建模方法,然后综合考虑选取造价影响因素,利用Python进行模型训练和预测。结果表明,GBDT预测效果要优于SVM算法,更适用于样本数据量偏少的LNG接收站造价预测。
LNG terminal project has the characteristics of high investment and long cycle,so it is very important to carry out accurate and efficient cost prediction in the early stage.This paper discusses how to use historical project data for machine learning to realize the project cost prediction of LNG terminal.Firstly,the SVM and GBDT modeling methods suitable for small sample data are determined through literature analysis,then chooses the cost influencing factors after comprehensive consideration,and Python is used for model training and prediction.The results show that,the prediction effect of GBDT is better than SVM,it is more suitable for the cost prediction of LNG terminal with le ss sample data.
作者
郑巧珍
张莹
徐双双
ZHENG Qiaozhen;ZHANG Ying;XU Shuangshuang(CNOOC Gas and Power Group Co.,Ltd,Beijing 100028,China)
出处
《建筑经济》
北大核心
2022年第S02期118-122,共5页
Construction Economy
关键词
LNG接收站
造价预测
梯度提升决策树
支持向量机
LNG terminal
cost prediction
gradient boost decision tree
support vector machine