An approach to integration of data mining with relational DBMS based on automatic SQL code generation

The paper introduces an approach to integration of data mining algorithms with relational DBMS. Approach assumes using XML-based markup for data mining algorithms implemented in SQL. User specifies input tables and algorithm parameters. After that system generates persistent stored procedures for specific data mining algorithm. Proposed markup language is described with examples. If there are several algorithms for some data mining problem the most effective algorithm implementation in SQL is generated by means of DBMS command EXPLAIN which shows estimated execution cost. Module structure and API for proposed program system is described.