数据误差对径向基网络的预测效果影响的理论研究  

Theoretical Study on the Influence of Data Error on the Prediction Effect of Radial Basis Function Network

作　　者：厍红科 樊炳怀 达朝究[1,2] SHE Hongke;FAN Binghuai;DA Chaojiu(School of Mathematics and Computer Science Institute,Northwest Minzu University;Institute of Applied Mathematics and Astronomical Calendar,Northwest Minzu University,Lanzhou,Gansu Province,730000 China)

机构地区：[1]西北民族大学数学与计算机科学学院 [2]西北民族大学天文历算研究所,甘肃兰州730000

出　　处：《大众科学》2024年第13期103-105,共3页China Public Science

摘　　要：以混沌系统Lorenz方程为研究对象进行探讨。在大数据背景下,误差是不可避免的,利用机器学习中的深度学习来研究,将Lorenz方程的数值解作为研究对象,首先以70%的训练集和30%的测试集进行试验,利用径向基神经网络进行预测,以此得出预测值与原始值的差距。其次在Lorenz方程中加入强迫项,利用模式误差进行试验,讨论并且计算了其误差等评价指标,波动程度较小。最终得出在加入误差之后的预测效果要比未加入的预测值效果好。In this paper,Lorenz equation of chaotic system is studied.In the context of big data,errors are inevitable.Deep learning in machine learning is used to study the numerical solution of Lorenz equation as the research object,and 70%training set and 30%test set are first tested,and radial basis neural network is used to make predictions,so as to obtain the difference between the predicted value and the original value.Secondly,the forced term is added to Lorenz equation,and the model error is used for experiments.The evaluation indexes such as the error are discussed and calculated,and the fluctuation degree is small.Finally,it is concluded that the prediction effect after adding error is better than the prediction effect without adding error.

关 键 词：数据误差 径向基神经网络 LORENZ 方程 算法设计 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]