Hi.
I'm rather new to the world of Business Intelligence, but am an experienced database developer, so I was plunged into this.
I need to make a prediction query using Analysis server and we're using the Association Rules for the prediction. I have a number of questions where I'm confused/unable to continue.

My model is that I have a table in a "datawarehouse" with Orders consisting of an Order ID and some additional information which in the future will become input variables for the prediction.
I then have an OrderLine table with a LineID, OrderID and ProductID.
Those products are associated with Product Groups (in a many-to-many relation ship in the source database) which is part of my problem, because I'm unsure how to design this model.

So...... getting to my problem:

I need to make a prediction query which takes Product ID as input and makes predictions on the associated Product ID's.
Now that's no big deal and I have that worked out to
Code:
SELECT FLATTENED PREDICT([My Nested Table], 5)
FROM [My Mining Model]
NATURAL PREDICTION JOIN 
(
 SELECT ( SELECT inputValues) AS [My Nested Table]
) AS InputTable
This works fine enough..... but

My problem is that I need to filter the result, so in some situations only Product ID's which are assigned specific Group IDs are returned from my prediction and in other situations I need to predict Product ID located in other Group ID's exclusive the one used as input.

I do not know if this is even possible as I can find no information anywhere to help me, which means I don't even know how to model the mining model, how to write the prediction query and so on.

I have full control over the datawarehouse and can make what ever models/datasource views etc in the analysis server. My only requirenment is high performance, but right now I am mostly interested in solving the problem and then worry about performance second.

Can anybody help me?