基于组合残差修正的优化神经网络预测方法  

Optimized Neural Network Prediction Method Based on Combined Residual Correction

作　　者：李博 廖梦洁[3] 张健 Li Bo;Liao Mengjie;Zhang Jian(Business School,Yangzhou University,Yangzhou Jiangsu 225127,China;School of Economics and Management,Beijing University of Technology,Beijing 100124,China;Beijing Key Laboratory of Big Data Decision Making for Green Development,Beijing 100192,China)

机构地区：[1]扬州大学商学院,江苏扬州225127 [2]北京工业大学经济管理学院,北京100124 [3]绿色发展大数据决策北京市重点实验室,北京100192

出　　处：《统计与决策》2025年第4期35-39,共5页Statistics & Decision

基　　金：北京市博士后工作经费资助项目(2024-zz-062);扬州大学高层次人才科研启动项目(137013600)。

摘　　要：为提高单一模型的预测精度,文章提出基于组合残差修正的优化神经网络预测方法。首先,基于麻雀搜索算法对神经网络参数进行优化,以避免预测精度降低;其次,引入IOWA算子求解神经网络模型的加权向量,以避免单一预测模型预测精度在不同时点时高时低;最后,对回声状态网络模型预测结果进行修正。为验证模型的有效性,以北京市猪肉月度价格为例进行实证分析,并与9个单一预测模型进行对比。结果表明:相较于其他单一预测模型,所提模型的预测精度更高,可以对其他类似特征的农产品价格进行准确预测。In order to improve the prediction accuracy of single models,this paper proposes an optimized neural network pre⁃diction method based on combined residual correction.Firstly,the sparrow search algorithm is employed to optimize the parame⁃ters of neural network to avoid a decrease in prediction accuracy.Secondly,the IOWA operator is introduced to calculate the weighted vector of the neural network model to avoid the fluctuating prediction accuracy of a single prediction model at different time points.Finally,the prediction results of the echo state network model are further corrected.In order to verify the validity of the model,the monthly pork price in Beijing is taken as an example for empirical analysis,and comparison is made with 9 single forecasting models.The results show that the proposed model has higher prediction accuracy than other single prediction models and can accurately predict the prices of other agricultural products with similar characteristics.

关 键 词：IOWA算子 神经网络 麻雀搜索算法 残差修正 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] F201[自动化与计算机技术—控制科学与工程]