Operation Data Processing And Analysis Of Mobile Communication Network

With the rapid development of information technology and the continuous expansion of the scale of mobile communication networks,the mobile network management center will receive a large number of alarm data all the time,which brings great burden to the network monitoring system.Alarm correlation analysis is an indispensable part of network management,which can mine related rules in alarm data to search root alarms related to faults.In addition,with the development of computer network,the security of communication networks has also attracted more and more attention.Intrusion detection technology is an important means to detect and prevent network threat behavior.Therefore,it is especially important for network managers to explore how to construct an effective mechanism to diagnose network failure and detect network intrusion.The main contents of this paper are as follows:(1)According to the implementation strategy of the existing alarm correlation analysis method and intrusion detection technology,these are classified and summarized.Specially,data mining algorithms widely used in alarm correlation analysis and intrusion detection are mainly introduced and analyzed.Next,the basic structure of mobile communication network and the related concepts of network fault and alarm are introduced.Then the network alarm data provided by mobile operator is presented from many aspects and the characteristics of alarm data are analyzed deeply.Last,the process structure of alarm data analysis is introduced and then the related process and processing results of alarm data preprocessing in this study are given.(2)A weighted alarm analysis method is designed.Considering the different effects of different alarms on the network in the communication network,the entropy method is firstly used to assign the corresponding weights for the different alarms,and then it is converted into sequence data set.Then a weighted alarm sequence pattern mining algorithm is designed to mine the weighted alarm sequence pattern in alarm data,and a novel pruning strategy is used to reduce the size of data set that needs to be searched,which improves the efficiency of the algorithm.Through the analysis of the alarm data of mobile communication network,the results show that the method has good performance in pruning effect,mining important alarm sequence patterns and executing efficiency.(3)A parallel anomaly detection algorithm based on Map Reduce framework is designed for network intrusion detection.In view of the shortcomings of the influence of repeated points on local outlier factor(LOF value)and high complexity of Local Outlier Factor(LOF)algorithm,the k-distance is redefined to avoid the situation where some LOF values of repeated points cannot be calculated.At the same time,the whole data set is logically divided into several data blocks,and the calculation of the k-distance and LOF value of each data point can only be performed in a single block by using the Map Reduce principle.Then the higher LOF values in each data block are merged and their LOF values are updated to obtain the final outliers.Finally,the results of anomaly detection of real network intrusion data show that this method has great advantages in accuracy,sensitivity and execution efficiency in detecting network intrusion anomaly data.