var sidebar_align = 'right';
var content_container_margin = parseInt('290px');
var sidebar_width = parseInt('270px');
Unanswered: Denormalization of tables for data mininig
Im developing a data mining application with java (netbeans) and db2 express-c aiming at an association rule analysis of a basket.
Therefore I need to denormalize tables in order to have every transaction in one row/tuple. That is to say I have to have multiple rows in one row as follows:
my transaction consists of a transaction ID (T-ID) and the related products (P-IDs) as well as the related employees (E-ID). Thus I have two normalized tables T1, T2 that should be denormalized:
T1: T-ID, P-ID > for the products being sold
T2: T-ID, E-ID > for the employees participating in the transaction
Lets suppose following contes:
To use a data mining algo i need to have these data in one row per transaction thus for the upper content there is just Yes/No for every attribute and every attribute is an attribute value:
Destination Table with these Attributes: T-ID, P1, P2, P3, E1, E2
Do you know wether there are kind of denormazilation algorithms/procedures in DB2 or Java to do so?
Last edited by digu; 09-29-11 at
I thought the way is not practical.
Because, if number of products and/or number of employees increased,
porcessing of the denormazed tables need to deal some issues, like
a) wheather processed by programs or human beings,
checking every columns of p1, p2, ..., e1, e2, ... may be repeated torublesome work.
b) the maximun number of columns in a table is restricted by DB2's limit(1012 on DB2 for LUW).
Anyway, if you want to denormalize like that way,
a procedure may be ...
1) join t1 and t2 by full outer join.
2) unpivot the joined table by using a technique like in
Last edited by tonkuma; 09-29-11 at
Reason: Add "in a table"