改进的小波神经网络在智能运维中的预测研究  

Prediction Research of Improved Wavelet Neural Network in Intelligent Operation and Maintenance

作　　者：周紫祥 周湘辉 王梦真 ZHOU Zixiang;ZHOU Xianghui;WANG Mengzhen(School of Mathematics and Statistics,Anhui Normal University,Wuhu 241003)

机构地区：[1]安徽师范大学数学与统计学院,芜湖241003

出　　处：《计算机与数字工程》2024年第7期2123-2128,2160,共7页Computer & Digital Engineering

基　　金：安徽省教育厅自然科学基金项目(编号:KJ2020A0055)资助。

摘　　要：随着大数据时代的到来,神经网络算法目前已经得到了很好的发展和应用,但在应用时,固定的学习率太大或太小都将面临收敛速度缓慢甚至发散的问题。因此,为避免经验因素对传统神经网络影响,文章在调整权值的基础上引入了自适应学习率修订函数的改进方法,以达到提升网络训练的速度和稳定性的目的。以小波神经网络为例,应用智能运维中的趋势预测问题进行仿真检验,结果表明相对于固定学习率,文章提出的自适应学习率改进算法能够有效提高小波神经网络的收敛速度,并有效降低了其收敛误差。With the advent of the era of big data,ANNS have been well developed and applied at present.However,in appli⁃cation,if the fixed learning rate is too large or small,it will face the problem of slow convergence and even divergence.Therefore,in order to avoid the influence of empirical factors on the traditional neural network,the paper introduces an improved method of adaptive learning rate revision function on the basis of adjusting the weight,so as to improve the training speed and stability.Taking the wavelet neural network as an example,it is applied to the trend prediction problem in intelligent operation and maintenance for simulation testing.The results show that compared with the fixed learning rate,the improved algorithm,which is based on a learn⁃ing rate that can be automatically adjusted,can effectively improve the convergence speed and reduce its convergence errors.

关 键 词：小波神经网络 自适应学习率 智能运维 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]