摘要
以氧化镥价格作为预测对象,选取2013年6月-2023年3月的月度数据,构建相关系数矩阵—非线性可逆网络模型(CM-INN模型),对氧化镥价格进行预测.预测结果表明,CM-INN模型能够解决异常值和离群点敏感、参数选择困难的问题,可有效处理复杂建模关系,提升特征值计算和模型训练效率.CM-INN仿真能力强、误差低、收敛精度高、计算效率高,对重稀土氧化镥价格走势有一定的预测意义.
Using the price of lutetium oxide as the prediction target,monthly data from June 2013 to March 2023 is selected to build a correlation coefficient matrix-nonlinear invertible network model(CM-INN model)for predicting the price of lutetium oxide.The prediction results show that the CM-INN model can solve the problems of sensitivity to outliers and abnormal values,and the difficulty in parameter selection.It can effectively handle complex modeling relationships,and improve the efficiency of eigenvalue calculation and model training.The CM-INN has strong simulation capability,low error,high convergence accuracy,and high computational efficiency,which has certain predictive significance for the trend of the price of the rare earth oxide lutetium.
作者
邓贞宙
赵旭
易金梅
DENG Zhenzhou;ZHAO XU;YI Jinmei(Rare Earth Research Institute,Nanchang University,Nanchang 340000,China;Teaching and Research Office of Economics and Management,Ganzhou Municipal Party School of the Communist Party of China,Ganzhou 341000,China)
出处
《牡丹江师范学院学报（自然科学版）》
2024年第3期1-5,共5页
Journal of Mudanjiang Normal University：Natural Sciences Edition
基金
国家自然科学基金项目(61501197)
国家自然科学基金优青培育项目(20202ZDB01002)
澳门青年学者计划项目(M201921)
关键词
多氧化镥
相关系数矩阵
非线性可逆网络
lutetium oxide
correlation coefficient matrix
nonlinear reversible network
