Three-Step method to tightly integrate data mining tasks into a relational database system

Abstract

In this paper, a result of the research project that aimed to define new algebraic operators and new SQL primitives for knowledge discovery in a tightly coupled architecture with a Relational Database Management System (RDBMS) is presented. In order to facilitate the tight coupling and to support the data mining tasks into the RDBMS engine, the three-step approach is proposed. In the first step, the relational algebra is extended with new algebraic operators to facilitate more expensive computationally processes of data mining tasks. In the next step and with the aim that the SQL language is relationally complete, these operators are defined as new primitives in the SELECT clause. In the last step, these primitives are unified into new SQL operator that runs a specific data mining task. Applying this method, new algebraic operators, new SQL primitives and  new SQL operators for association and classification tasks were defined and were implemented into the PostgreSQL DBMS engine,  giving it the capacity to discover  association and classification rules efficiently.