I have bad performance in my Basket Analysis (Association Rules) Mining
Cubes. The size is 12milj fact rows, and 5 regular dimensions of 1 000 - 25
000 members and 4 Datamining dimensions, with about 15 - 3000 members, but
there are about 8 milj baskets at the lowest level in these DM-dimensions.

To open my mining cube with a filter on one of the datamining dimensions,
takes about 3-8 minutes.

I have set up aggregations on the original cube, but what can I do to do
speed up the queries when filtering with the Mining dimensions? (I belive
that the Mining dimensions acts like Many to Many relations with the fact
table)

The Linked cube doesn't allow aggrigations (as far as I know)

Best Regards, Joakim

The parameters of associaion rules mining are maybe the reason of bad
performance.

You can try to set higher values of parameters.
Like set minimum_probability as 0.5, and set minimum_support as 0.1 even
more.

The default value often causes bad perfromance.

Best Regards,

OK, good point!
But I've already set these parameters to support the dependices network
screen and other requirements. My values are Min_Prob = 0.1 and Min_support
to 1000
these settings gives me between 400 and 1800 (to be more exact than my
earlier description) members in my DM-dimensions.

Is there any other way to preaggregate or do something else to get better
performance?

//Joakim

"ivan wang" wrote: