Design And Implementation Of Alarm Rule Mining System In Communication Network

In the era of popularization of communications and networks,people have high requirements for the quality of communications networks.The failure of the communication network equipment will not only lead to a decrease in network quality,but also have a negative impact on the user experience,and even cause economic losses.As a result,the requirements for the quality of the communication network have reached a harsh level.Communication network alarms are the most sensitive signals that reflect the state of the communication network.The current communication network is dense and complex,and failures caused by a single cause can easily cause an alarm storm.The existing analysis logic is: give rules in advance to perform alarm merge and analysis.This method has no way to automatically discover new rules and cannot respond to emerging situations;it cannot effectively cope with the increasingly complex networking structure of the communication network and deal with alarm storms.Therefore,the automatic discovery of communication network alarm rules is valuable and meaningful.This paper analyzes the characteristics of the alarm data of the communication network and analyzes it with the association analysis algorithm,so that the alarm rules can be actively discovered.According to the characteristics of the actual alarm data and the requirements of the data structure required for the algorithm input,the design and implementation of extracting alarm data from the existing network,associating resources,filtering invalid values,performing effective grouping and other operations to meet the input requirements of the data mining algorithm.This article adopts batch processing to distributed,and then to streaming data processing methods;applying big data and streaming data processing analysis platforms to mine relevant alarm rules.Analyze,classify and summarize the results mined by the algorithm.Complete the design,implementation and functional testing of the interface and data mining processing related modules.The main achievements are: using the idea of software engineering,combined with the actual analysis needs of the alarm data of the communication network,and applying the relevant algorithm technology of correlation analysis in data mining,the system needs analysis,detailed design and implementation,and finally gives the test results and analysis,summary,induction.Design and implement the relevant modules of the interface layer to realize the functions of viewing results,sending tasks,and dispatching expert groups.Realize the application of frequent mode algorithm(Apriori,FP-Growth),sequence mode data mining algorithm(Prefix Span)to analyze the alarm data,and obtain the available alarm frequent mode and alarm sequence mode.Design and implement related modules from distributed dataanalysis to distributed analysis based on machine learning(MLlib)package on Spark cluster to realize analysis of massive alarm data.Design and implement related modules from static to streaming data analysis(Spark Streaming)to realize the analysis of alarm data flow.Analyze the frequent patterns and sequence patterns in the alarm data of the communication network,analyze and classify the excavated alarm rules,and explore the parameter reference thresholds that can be used in practical applications.And the system is fully tested to achieve the desired effect.