Hey mates I am trying to implement an association rule mining in a vertica db: So far I got the heavy lifting done by finding the frequent itemsets however I still need to generate the rules. Here is an example:
Initial table:
+--------+-------+-------+
| item1 | item2 | item3 |
+--------+-------+-------+
| A | B | C |
+--------+-------+-------+
With 3 items I can generate 6 rules, a rule has a rule body and a rule head. This would result in the following table:
+---------+---------+--------+--------+
| ItemB1 | ItemB2 | ItemH1 | ItemH2 |
+---------+---------+--------+--------+
| A | Null | B | C |
| B | Null | A | C |
| C | Null | A | B |
| A | B | C | Null |
| A | C | B | Null |
| B | C | A | Null |
+---------+---------+--------+--------+
This is a very abstract example, in reality there is more than one item set and more than 3 items in one item set. I found some papers that discuss association rule mining and claim that this part is trivial. Guess not so trivial to me.
Thanks
