基于二阶段异质随机森林的汽油辛烷值预测  

Prediction of Gasoline Octane Number Based on Two-stage Heterogeneous Random Forest

作　　者：谢鑫 王胜 晏涛 王旋晔 XIE Xin;WANG Sheng;YAN Tao;WANG Xuanye(School of Mathematical Science,Sichuan Normal University,Chengdu,Sichuan 610066,China)

机构地区：[1]四川师范大学数学科学学院,四川成都610066

出　　处：《数学建模及其应用》2021年第1期39-44,共6页Mathematical Modeling and Its Applications

摘　　要：构建了基于二阶段异质随机森林的汽油辛烷值预测模型.首先利用样本-位点信息表知识约简模型,筛选出对汽油辛烷值影响大的位点数据作为第一阶段;然后,利用集成学习思想集成支持向量回归和动态时间序列神经网络,构建异质随机森林预测模型作为第二阶段.利用十折交叉法验证模型精度,结果表明该集成学习算法具有有效性和高精度.In this paper,a gasoline octane number prediction model based on two-stage heterogeneous random forest is constructed.Firstly,the knowledge reduction model of sample site information table is used to select the site data which has great influence on octane number as the first stage;then,the heterogeneous random forest prediction model is constructed by integrating SVR and dynamic time series neural network by ensemble learning idea.The results show that the ensemble learning algorithm is effective and accurate.

关 键 词：集成学习 信息熵 知识约简 支持向量回归 动态时间序列神经网络 

分 类 号：O151[理学—数学]