Applied Research Of Parallel Data Mining Techniques In Telecommunications Network Management Alarms

With the continued development of the telecommunications network technology, telecommunications networks generate heterogeneous and dynamic resources, resulting in network management system functions more and more sophisticated and complex. Telecommunications network system will generate real-time massive alarm data, usually a carding process alarm management system detects a fault triggered the alarm, the alarm information is presented to the network administrator. Due to the generation of a fault can lead to different devices have different multiple alarms, but not every information generated alarm can describe the root cause of the fault, the fault alarm management system for accurate locating brings challenges. Traditional network management systems and network management personnel to deal failure diagnosis, location and recovery based solely on their own experience and systems limited functionality. With the growing size of the telecommunications networks, personal knowledge and experience has been unable to meet the needs of the telecommunications network alarm processing.Data mining provides a new method for telecommunications network alarm analysis. Currently, the telecommunications network is developing to the cluster architecture, data mining algorithm parallelism based on Hadoop’s MapReduce model will further optimize the performance of massive alarm data analysis.In this paper, the telecommunication network system fault alarm correlation is outlined, introduced the concept of telecommunications network failures and alarms; Research of alarm correlation analysis methods; discusses the specific application of data mining algorithms in the network system firstly. Secondly, give an owerview of Hadoop Platform; research and analysis the structure and operation of MapReduce programming model; Then Apriori algorithm is discussed in detail in its Hadoop platform parallelization proposed variable multi-stage parallel algorithm VMSPA, depending on the comparative with number of frequent itemsets between two adjacent iterations to dynamic switches simple parallel algorithm and VMSPA. VMSPA algorithm dynamically updates the number of candidate itemsets generated simultaneously to avoid frequent scanning the transaction database, thereby the performance of the algorithm is enhanced. Finally, the variable multi-stage parallel algorithms is used in telecommunications network alarm correlation analysis.After the performance of the algorithm is analyzed by experiments the effectiveness of the algorithm is verified, By applying the algorithm in the alarm correlation analysis to prove that the method has some practical value.