Research On Alarm Prediction Of Mobile Communication Network Based On Weighted Association Rules

At present,people often come into contact with mobile communication technology in their daily work and life.With the increasing scale and complexity of mobile communication network,there are more and more communication network failures.Failures will lead to a series of equipment alarms,and it is particularly important to predict the upcoming failure alarms of communication network equipment through the alarms that have occurred.Alarm association rules can establish the association relationship between alarms for alarm management and prediction.Based on the weighted association rules of communication network alarms,this paper conducts prediction research on the possible alarms in the future.The main work is as follows:1)By comparing the existing alarm correlation analysis methods,and the alarm correlation mining method based on data mining is determined for alarm mining.According to the characteristics of mobile communication network alarms and the data requirements of the determined mining methods,the alarm data preprocessing method and process are determined.2)By comparing the relative importance of the alarm level with other attributes of the alarm,alarm analysis tree of communication network is established.Analytic Hierarchy Process(AHP)is used to compare the relative importance of alarm level and the maximum number of branches of the node where the alarm is located.The relative weight and final weight are calculated,which are used to assign weights to the preprocessed alarm sequence.3)The classic alarm association rules Apriori and FP-growth are analyzed,and a weighted FP-growth alarm association rule mining method is proposed.The weighted conditional FP-tree is constructed to mine the alarm association rules,and the alarm transaction database is traverses.A large number of alarm item sets less than the weighted support are effectively filtered to generate the effective strong alarm association rules.By mining the real alarm data of the mobile communication network,it is verified that the association rule mining method is superior to FP-growth in terms of pruning effect and frequent patterns mining.4)The alarm prediction model of communication network based on pattern matching is established.The real-time alarm sequence is input,and the alarm prediction model is used to predict the alarm.The experimental comparison shows that the proposed method is superior to the alarm prediction based on FP growth in both execution efficiency and prediction efficiency.