基于特征选择与梯度提升决策树的球团矿抗压强度预测
Prediction of compressive strength of pellet based on feature selection and gradient boosting decision tree
陈许玲 1蒋文浩 1黄晓贤 1范晓慧 1甘敏 1曹风1
作者信息
- 1. 中南大学资源加工与生物工程学院,湖南 长沙 410083
- 折叠
摘要
抗压强度是评价球团矿质量的重要指标,也是球团生产的核心控制目标,但其检测周期长、控制严重滞后.因此,实时准确预测球团矿抗压强度对提升和稳定球团质量具有重要意义.提出一种基于 Filter 和 Wrap-per混合特征参数遴选结合贝叶斯优化算法的球团矿抗压强度预测方法,采用现场生产数据对模型进行训练与测试,预测结果表明:特征选择和贝叶斯优化算法的引入可以明显提升模型的预测精度,基于特征选择和贝叶斯优化的梯度提升决策树(GBDT)模型的拟合效果最好,预测准确率达 95.31%,为球团矿质量的优化控制奠定了良好的基础.
Abstract
Compressive strength is an important indicator for evaluating the quality of pellet ore,and it is also the core control objective of pellet production.However,its detection cycle is long and the control is seriously lagging behind.Therefore,accurate real-time prediction of pellet compressive strength is of great significance for improving and stabilizing pellet quality.A prediction method of pellet compressive strength based on Filter and Wrapper mixed feature parameter selection combined with Bayesian optimization algorithm was proposed.Field production data was used to train and test the model.The prediction results show that:feature selection and the introduction of Bayesian optimization algorithm can significantly improve the prediction accuracy of the model.The gradient Boosting decision tree(GBDT)model based on feature selection and Bayesian optimization has the best fitting effect,with the prediction accuracy of 95.31%,laying a good foundation for the optimization and control of pellet quality.
关键词
球团矿抗压强度/梯度提升决策树/特征选择/模型预测Key words
compressive strength of pellet/gradient boosting decision tree/feature selection/model prediction引用本文复制引用
出版年
2024
NETL
NSTL
万方数据