A Seamless Integration of Association Rule Mining with Database Systems

TL;DR

This paper presents a unified approach to integrating association rule mining with database systems using a common query optimizer, enabling flexible, constrained queries and improved user control in data mining tasks.

Contribution

It introduces a seamless integration scheme employing a common query optimizer and algebraic query trees, enhancing flexibility and user control in association rule mining within databases.

Findings

Unified query optimizer for data mining and database queries

Flexible algebraic representation for constrained association rules

Modular query tree simplification for complex data mining tasks

Abstract

The need for Knowledge and Data Discovery Management Systems (KDDMS) that support ad hoc data mining queries has been long recognized. A significant amount of research has gone into building tightly coupled systems that integrate association rule mining with database systems. In this paper, we describe a seamless integration scheme for database queries and association rule discovery using a common query optimizer for both. Query trees of expressions in an extended algebra are used for internal representation in the optimizer. The algebraic representation is flexible enough to deal with constrained association rule queries and other variations of association rule specifications. We propose modularization to simplify the query tree for complex tasks in data mining. It paves the way for making use of existing algorithms for constructing query plans in the optimization process. How the integration scheme we present will facilitate greater user control over the data mining process is also discussed. The work described in this paper forms part of a larger project for fully integrating data mining with database management.