Network Alarm Correlation Based On Sequential Pattern Mining

Data Mining is the procedure of extracting and mining knowledge from large amount of data in database and other information repository. It is the integration of multiple subjects of technology and it's one of the hottest research areas currently. In the real world with large amounts of data on the timing or order of relevant, sequential pattern mining is looking for a certain time window constraints with some attributes of alarm events have led other attributes with the incident will be warning in another probability of a time window within the rules.This paper figures out a comprehensive study on the sequence pattern, discusses the research in the field of the status, the latest technology and progress, analyses the six types of mining sequential patterns algorithms, which also raised the algorithms' advantages, disadvantages and application environment.Meanwhile, Author analyzed the existing communication network alarm data mining and gave us the overall solution. The author innovatively combines the network topology constraints and sequential pattern mining to make two new algorithms which are based on the large amount of network elements and data.In addition, the author developed BUPTPRISMiner data mining platform with the team members and integrated the sequential pattern mining algorithms into the platform, and through experiments prove that, with the network topology constraints, the sequence mining algorithms cost less time, efficiency substantially improved the speed and accuracy.