Clustering Dimension

I am new to data mining and have a question about OLAP dimensions built from models. Do you know if you can use a dimension that was created by the mining model wizard in the same cube that is being used as the source for mining model it self? I keep getting an error about a dependency loop and just want to make sure that I am not trying to do something impossible.
If this is illegal, do you know of a way of doing this without essentially having one cube dependant on another via a mining model and dimension? I tried to use the wizard to create a mining off of a table but it did not give me the option of creating a dimension from it.

You have to create a new cube that is "essentially" dependant on the other via the model and dimension if you want to use the model directly as a data mining dimension. This uses linked measure groups and doesn't really impact performance or storage space - it's mostly just metadata.

The only option of using a table-based model as a dimension is to write out the cluster labels and simply make the cluster label as a dimension attribute. You could even append the cluster label to the source data (e.g. the customer table) and not have a seperate dimension, simply a browseable attribute on the dimension of interest

|||Makes sense. Thanks.|||

Jamie, can you provide more information on how to do this? We'd like to have a series of clusters in an existing household dimension. That is, we need multiple occurences of cluster model results. I've looked at the data source, dimension, and cube created by the data mining model, but I don't see where the case ID (Household Key) and the cluster name could be extracted to update the existing dimension. We're using the cube for the data mining source.

I'd also like to know how to keep the linked cube and the source cube metadata in sync. If I make a change to the source cube, say by adding a new measure, the metadata for the linked cube gets out of sync. I've been deleting the data mining dimension, cube, and dsv and them adding them back in using the data mining menu in the model. What am I doing wrong here?