Network Fault Localization Based On Network Topology And Data Mining

Based on the project developed cooperatively with enterprise:《Key Techniques in Intelligent Mobile Network Fault Management based on Data Mining》, the paper focused on the application of techniques of mining association rules in the mobile network fault localization system.The project reaches a conclusion that alarms spread through network topology, after studying the mobile network and its alarms. Based on the conclusion, the paper presents introducing network topology in mining association rules to increase mining efficiency and validity. The network topology can be used to process the alarm data during mining, thus filtrating alarms and association rules that have no topology correlation.In allusion to the deficiency of traditional algorithms for mining association rules, the paper presents using FP-Growth algorithm and brings network topology into the algorithm. This algorithm scans the database only twice, is best for huge database, can save time and improve execution efficiency.The author participated in the study and development of the BUPTPRISMiner intelligent alarm analysis system. Concretely, the author took part in realizing the FP-Growth algorithm, help other team members to integrate the algorithms into the BUPTPRISMiner system. Meanwhile the author accomplished the module testing, code maintenance, document writing and so on.