Bi-Apriori-Based Association Discovery via Alarm Logs

Abstract

The state grid information system is complex, the operation and maintenance information are diverse, involving a wide range of aspects. It becomes a key problem that how to use the alarm log of the operation and maintenance system to analyze the root cause of the fault. At present, there is a lack of alarm correlation technology for operation and maintenance system, and the traditional method lacks mature application systems for new scenarios. The reliability and efficiency of operation and maintenance system not only depends on the accurate collection of equipment fault information, but also depends on how to analyze massive fault information effectively and quickly, so as to grasp the key, handle the key, solve the key, three “keys”. In this context, a Bi-Apriori algorithm is proposed, to solve the problem of frequently scanning the dataset and generating a large number of candidate set. It adopts the vertical data structure to mine frequent items and then as for alarm logs the proposed Bi-Apriori can be used to mining frequent items from alarm logs. Experimental results show that the proposed Bi-Apriori algorithm is superior to the existing association rule discovery algorithms on several datasets in real applications, especially for the dataset of alarm logs.