一种基于BPNN的智能巡检异常预测模型的研究  

A Study on a BPNN-Based Anomaly Prediction Model for Intelligent Inspection

作　　者：邓小清 王伦浪 王刚 蒲国林 DENG Xiao-qing;WANG Lun-lang;WANG Gang;PU Guo-lin(School of Intelligent Manufacturing,Sichuan University of Arts and Science,Dazhou Sichuan 635000,China;School of Electronic and Information Engineering,Ankang University,Ankang Shaanxi 725000,China)

机构地区：[1]四川文理学院智能制造学院,四川达州635000 [2]安康学院电子与信息工程学院,陕西安康725000

出　　处：《西南大学学报（自然科学版）》2019年第10期142-148,共7页Journal of Southwest University(Natural Science Edition)

基　　金：国家自然科学基金项目(61152003);四川省教育厅重点项目(16ZA03532016);四川文理学院重点项目(2016KZ002Z)

摘　　要：在传统网络巡检方法中,网络异常发现主要基于单一参数进行阈值触发,误报率较高,效率低.为了高效准确地发现网络异常,提出了一种基于BPNN的网络异常预测模型.首先对采集系统采集的数据进行特征提取和初始化处理;然后,将初始化后的数据作为神经网络样本进行训练,根据误差阈值调整网络参数,确定网络结构;最后,在Matlab环境下进行仿真实验,将提出的BP神经网络模型用于网络异常预测,结果表明本文提出的方法对网络异常预测有较高的预测率.In traditional network inspection methods,the network anomaly discovery in intelligent patrol inspection is triggered by a single threshold based on performance parameters,with high false positive rate and low efficiency.In order to predict network anomalies effectively and accurately,we propose a network anomaly prediction model based on BPNN (BP neural network)in this paper.The collected data are extracted,preprocessed and trained as neural network samples,and then the network parameters are adjusted according to the error threshold to determine the network structure.A simulation experiment is carried out in Matlab environment,and the proposed BP neural network model is used for network anomaly prediction.The results show that the method proposed in this paper has satisfactory prediction rate for network anomaly prediction.

关 键 词：BPNN 智能巡检 异常预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]