Study of alarm association rules mining based on matrix mapping

The efficiency of the mining algorithms plays an important role in tackling with large datasets. Classical algorithms need to scan database many times likes Apriori, for it is difficult to mine important rules in telecommunication networks. This paper proposes a matrix mapping based association rules mining method MM-ARM. Compared with Apriori algorithm, MM-ARM needs to visit database only once. By scanning the database and applying a matrix mapping technique at the same time, the algorithm converts the alarm database into matrix, and save them in FP-tree structure in order to avoid scanning database repeatedly. The performance testing of the algorithm indicates that MM-ARM has higher temporal efficiency, which is suitable for mining association rules in telecommunication networks.