Multi-level Frequent Pattern Mining on Pipeline Incident Data

Connor C. J. Hryhoruk, Carson K. Leung, Jingyuan Li, Brandon A. Narine, Felix Wedel

Abstract: Pipeline incidents occur throughout the world and can have devastating financial and ecological impacts. A large amount of data is collected and made publicly available for each pipeline incident. Although some important information is explicitly visible in such datasets, a lot of implicit information remains hidden. In this paper, we explore frequent pattern mining approaches—specifically, multi-level frequent pattern mining, which help discover some of this implicit information that was previously hidden in the data. The resulting frequent patterns and their corresponding association rules provides some new insights into pipeline incidents that may help improve the safety and reliability of pipelines. Evaluation results show the effectiveness and practicality of our multi-level frequent pattern mining solution.