Research And Implementation Of Mobile Communication Network Alarm

With the rapid development of mobile communication network technology and the continuous expansion of the network scale,the network is becoming more and more complex,which makes the mobile communication network more and more difficult to manage.In case of network fault,it will cause a series of equipment to generate a lot of alarms,and even cause service interruption,which will bring great cost to operators.Therefore,it is very important to predict the alarm accurately and take measures in advance to reduce the occurrence of network fault.Traditional network fault diagnosis mainly relies on the experience of network managers,which is subjective and inefficient.With the rapid development of data mining technology,network alarm prediction based on data mining has gradually become a research hotspot,which has important practical value.This paper investigates the problem of mobile communication network alarm prediction based on data mining.The research and innovation mainly include:Firstly,an evaluation index called value function v is proposed to evaluate the performance of the network alarm prediction problem.Secondly,considering the time correlation of alarm data,a multi-class network alarm prediction algorithm called LM-NAP based on the Long Short-Term Memory(LSTM)is proposed,which is compared with the algorithms of simple neural network,Random Forest and Adaboost.The result shows that the LM-NAP algorithm has better performance in terms of value function v and accuracy.Thirdly,a network alarm association analysis method(CSWA)is proposed,including the clustering of base stations,alarm transaction generation,alarm importance quantification analysis,as well as weighted alarm association analysis.Finally,in order to improve the performance of LM-NAP,a two-class network alarm prediction algorithm(LAB-NAP)based on LSTM and association rules is proposed.Compared with the two-class network alarm prediction algorithm(LB-NAP)based on LSTM and the LM-NAP algorithm,the result shows that the LAB-NAP algorithm has better performance in practical network scenarios with large amounts of data and many types of alarms.This paper not only provides theoretical support for network fault management of operators,but also provides reference ideas for the practical application of data mining in network alarm analysis.