There's been several good posts on using the node description of a model as the end user description for a specific cluster. My model uses a number of continuous input columns defined as currency from a fact table in the source cube. After processing, the node description has elements that look like this:
Since the source data is currency, this makes the node description look a little strange. The data type in the model is set as double. The precision implied by the description is not what I want the model to consider. In the case above, the difference between the numbers listed is not significant.
It would be great to have a better node desciption that doesn't imply so much precision, but the bigger question is why does the cluster model turn currency types into doubles. Should I set the data type to long in the model so that cents are ignored? I know I should probably use discrete inputs, but I don't want to have to discretize the currency values in the cube since this would require me to set up fact dimensions for each currency column in the fact table.
Sorry, this is a limitation in the data mining engine. Changing the DM type in the mining structure to Long is the right workaround if the fractional values are not significant for the model.
No comments:
Post a Comment