基于改进GRU-DNN的收货风险预警模型构建及仿真

Construction and Simulation of Risk Early Warning Model for Receiving Products Based on Improved GRU-DNN

陈清兵 ¹章光东 ¹徐康 ²肖志敏³

扫码查看

作者信息

1. 国网安徽省电力有限公司,安徽,合肥 230022
2. 国网安徽省电力有限公司物资分公司,安徽,合肥 230061
3. 国网安徽省电力有限公司池州供电公司,安徽,合肥 247100
折叠

摘要

针对现有收货风险预警方法准确率低的问题,结合Gate Recurrent Unit(GRU)和Deep Neural Networks(DNN),提出一种基于改进GRU-DNN模型的收货风险分析方法.通过采用模糊综合分析法(FSA)筛选出收货风险主要评价指标,并将指标输入通过对抗性训练与预测相似性改进的GRU-DNN网络中进行分类识别,实现了收货风险分析.仿真结果表明,所提的改进GRU-DNN风险预警方法可实现收货风险预警,且在准确率、精确率、召回率、F1各项指标上表现良好,均达到86％以上的有效率,相较于传统基于 DNN、Convolutional Neural Network(CNN)和 Multivariable Linear Regression Model(MLR)等模型的风险预警方法,具有明显的优势和良好的预测性能和鲁棒性.

Abstract

In view of the low accuracy of the existing receiving risk early warning methods,a receiving risk analysis method based on the improved GRU-DNN model is proposed.By using fuzzy comprehensive analysis(FSA)to screen out the main evaluation indicators of receiving risk,and input the indicators into GRU-DNN network improved by adversarial training and prediction similarity for classification and identification,the receiving risk analysis is realized.The simulation results show that the improved GRU-DNN risk early warning method can realize the receipt risk early warning,and performs well in accuracy,precision,recall and F,indicators,with an efficiency of more than 86％.Compared with the traditional risk early warning methods based on DNN,CNN and MLR models,it has obvious advantages and good prediction performance and robustness.

关键词

风险评价/模糊聚类/GRU模型/DNN网络

Key words

risk assessment/fuzzy clustering/GRU model/DNN network

引用本文复制引用

出版年

2024

微型电脑应用

上海市微型电脑应用学会

微型电脑应用

CSTPCD

影响因子：0.359

ISSN：1007-757X

段落导航