An alarm data mining method for operation and maintenance support network based on stream computing

Abstract

Communication system failure will produce a large number of alarm data. If it cannot be mined on time, it will affect the normal operation of the whole system. Based on this, an alarm data mining method of operation and maintenance support network based on streaming computing is proposed. The discrete alarm data is transformed into alarm transactions suitable for mining by sliding time window algorithm, and the transaction weight value is calculated by analytic hierarchy process; Spark framework and MapReduce framework are adopted to process data in batch, parallel and real-time, the alarm data mining of operation and maintenance support network is realised by setting a reasonable alarm event database threshold. Comparative experimental results show that the proposed method occupies the smallest memory and the shortest running time, and the performance of alarm data mining is the best.