融合模糊预测控制算法与神经网络算法的数字化信息系统研究  

Research on digital information system integrating fuzzy predictive control algorithm and neural network algorithm

作　　者：杨丹 YANG Dan(The open University of Shaanxi,Xi’an 710119,China;Shaanxi Business College,Xi’an 710119,China)

机构地区：[1]陕西开放大学,西安710119 [2]陕西工商职业学院,西安710119

出　　处：《自动化与仪器仪表》2025年第3期201-205,共5页Automation & Instrumentation

基　　金：陕西工商职业学院2020年度科研重点课题《基于中高职贯通的招生模式研究》(20GA07)。

摘　　要：针对高校网站数据流量预测准确率低的问题,提出一种模糊预测控制算法与神经网络算法相结合的高校网站数据流量预测方法。首先,以数字化信息系统为研究对象,采用长短期记忆神经网络(Long Short-Term Memory,LSTM)作为网络流量基础预测模型;然后通过差分进化算法(Differential Evolution Algorithm,DE)对LSTM神经网络的隐含层节点和学习率进行优化;最后结合季节性差分自回归移动平均模型(Seasonal Autoregressive Integrated Moving Average Model,SARIMA)的非线性特征拟合性能,将其与DE-LSTM模型进行组合预测,最终得到基于SARIMA-DE-LSTM的组合预测模型。结果表明,SARIMA-DE-LSTM组合预测模型在高校数据流量预测工作中的RMSE误差、MAE误差和MAPE误差分别取值为0.482 3、0.517 1和0.363 9,明显低于BP神经网络预测模型和CNN-BiLSTM预测模型。由此说明,本模型能够提高高校网站流量预测精度,具备较高的预测性能,满足数字化信息系统的数据处理和流量预测需求。Aiming at the low accuracy of university website data traffic prediction,a kind of university website data traffic prediction method combining fuzzy prediction control algorithm and neural network algorithm is proposed.first,With the digital information system as the research object,Using a long-and short-term memory neural network(Long Short-Term Memory,LSTM) as the network traffic base prediction model;Then,through the differential evolution algorithm(Differential Evolution Algorithm,DE) Optimize the hidden layer nodes and learning rate of the LSTM neural network;Finally,combined with the seasonal differential autoregressive moving average model(Seasonal Autoregressive Integrated Moving Average Model,SARIMA) of the nonlinear feature fitting performance,Combined combined prediction with the DE-LSTM model,Finally,the combined prediction model based on SARIMA-DE-LSTM was obtained.The results show that the RMSE error,MAE error and MAPE error of the SARIMA-DE-LSTM combined prediction model in university data flow are 0.482 3,0.517 1 and 0.363 9,respectively,which are significantly lower than the BP neural network prediction model and CNN-BiLSTM prediction model.This shows that this model can improve the accuracy of university website traffic prediction,have high prediction performance,and meet the requirements of data processing and traffic prediction of digital information system.

关 键 词：长短期记忆神经网络 数字化信息系统 差分进化算法 SARIMA 流量预测 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]