Extraction Method of Alarm Transaction Based on Morphology Similarity Clustering
Shaoguang Liu, Jun Xie, Zhicheng Zhao, Yang Li, Xin Yan Yang · 2019
In analysis of power communication alarm correlation, the original relational alarm data needs to be converted into alarm transactional data for mining association rule. Considering the low efficiency of traditional sliding time window that extracts alarm transactions, this paper proposes a method of extracting alarm transactions based on morphological similarity clustering. Considering spatial correlation of power communication alarms, the alarm sequence is divided into numerous sequence groups, which then are clustered into many clusters by morphological similarity measure. In each cluster, different time window width and sliding step are set to extract the alarm transactions. Experiments show that the time window method based on morphological similarity clustering has higher efficiency of extracting alarm transactions and is beneficial to acquisition of root fault.