Research On Classification Of Network Alarm Information Based On Data Mining

With the rapid development of Internet,the scale of network system has become more and more complex and huge.How to ensure its smooth operation has become an important part of network operation and maintenance.Network alarm is an important way to operate and maintain the network system.When the network system has the risk of failure,the system will send an alarm to remind the operation and maintenance personnel to check it.At the same time,the increasing complexity of the network system will lead to a large number of alarms,which will not only bring huge troubleshooting pressure to the operation and maintenance personnel,but also lead to the risk of missing core alarms.The alarm troubleshooting of the network system is the key to the continuous stability of the operation and maintenance network system,which requires us to accurately classify the alarm information to improve the alarm accuracy and reduce the troubleshooting pressure.The research of this paper is based on this.Firstly,we extract the features of various heterogeneous alarm attributes.In this paper,we use the combination method of search and attribute screening to extract the features.In the process of search,we use rank rule to select the features in order.In the process of feature selection,we use infotain attributeeval method to evaluate and select the attributes.Then we use Bayesian network classification algorithm and Bayesian network classification algorithm based on full feature and feature extraction respectively The classification model of multilayer feedforward artificial neural network is established.Then,the performance of the classifier is compared from the aspects of classification accuracy and model training time.At the same time,the modeling effect of different feature selection is compared with the horizontal dimension of feature selection,and the classification table of different classification algorithms under the same feature subset is compared with the vertical dimension of classification algorithm.Finally,according to the experimental results and the actual application of network alarm,a combined classification algorithm is constructed.Combined classification method combines Bayesian network classification and multilayer feedforward neural network classification,solves the problems of low classification accuracy of Bayesian network and long time consuming of multilayer feedforward neural network classification model,which can classify alarm information more accurately and eliminate invalid alarm information better.