QUARRY: A Graph Model for Queryable Association Rules

Association rule mining is a pivotal technique for knowledge discovery, but often involves time-intensive manual labour when performed on large datasets. In this paper we propose a solution for this problem: QUARRY, a graph model that enables consumable and queryable insights from association rules. In contrast to existing systems which take a list of rules and display them in a purpose-built visualisation, our graph-based model enables association rules to be queried directly via graph queries. Through a case study on maintenance data we show how this model enhances knowledge discovery by eliminating the need for domain experts to trawl through large lists of rules to find useful information. QUARRY, which is designed for compatibility with existing knowledge graphs, provides users with the means to easily search for rules pertaining to specific items as well as roll up and drill down on their searches using the concept hierarchy. Domain experts may also query for association rules based on transaction properties such as costs and dates, enabling critical insights into their data.