Exploratory Mining via Constrained Frequent Set Queries

Although there have been many studies on data mining, to date there have been few research prototypes or commercial systems supporting comprehensive query-driven mining, which encourages interactive exploration of the data. Our thesis is that constraint constructs and the optimization they induce play a pivotal role in mining queries, thus substantially enhancing the usefulness and performance of the mining system. This is based on the analogy of declarative query languages like SQL and query optimization which have made relational databases so successful. To this end, our proposed demo is not yet another data mining system, but of a new paradigm in data mining - mining with constraints, as the important first step towards supporting ad-hoc mining in DBMS. In this demo, we will show a prototype exploratory mining system that implements constraint-based mining query optimization methods proposed in [5]. We will demonstrate how a user can interact with the system for exploratory data mining and how efficiently the system may execute optimized data mining queries. The prototype system will include sill the constraint pushing techniques for mining association rules outlined in [5], and will include additional capabilities for mining other kinds of rules for which the computation of constrained frequent sets forms the core first step.