SQL-based Frequent Itemset Mining

  Overview
This page contains materials for mining frequent itemsets by sql queries. For details, see the paper "Shaping SQL-Based Frequent Pattern Mining Algorithms".
For more information contact scs@elte.hu.
  Test datasets
The public datasets we used can be reached on the site of the FIMI Workshop.
  Implementations
You can find three different PL/SQL codes here.
The algorithms we used are implemented and tested as PL/SQL procedures, with Oracel 9.2. The zip file contains *_init.sql files for creating DB objects initially, and other .sql files for creating packages. The fim_common package contains - mostly not used - procedures for testing. We eliminated most of the database user-specific and the testing purpose code parts. Both three algorithms can be used similarly, for example by the

begin
   fp_tdg2.mine(table_name =>'ACCIDENTS' , tid_attr =>'TID' , item_attr =>'ITEM' ,
   minsup_perc => 90 ,test_name => 'fp_tdg2', test_descr => 'testing 2') ;
end;

statement (with input schema, minimum support and test description parameters). The results can be found in the fp_tdg2_result table, or in the subquery_f_* tables in case of the subquery algorithm.