Thursday, March 29, 2012

Error: Decision Trees found no splits for model

Hi,

I am trying to run one of the mining models from the book "Delivering BI using SQl Server 2005" but I am running into "Decision Trees found no splits for model". The mining structure has 4 columns, the fourth one being marked as "Predict Only". My Cube slice for the model has sufficient data in the cube. I am lost.. Help!!

Regards

I'm not familiar with the DM section of that book. Could you give more descriptions of the data and the problem? What does the root node of the tree look like? Are you predicting a continuous or discrete value? Etc.

If you want to see the data the data mining is using, turn Drillthrough on for the model and then do

SELECT * FROM <model name>.CASES

To see the cases that were used for training the model. If this shows anything odd, you may see what's wrong.

|||

The mining model was created using Customer Dimension and Sales Fact. Following are the details of the mining model

Data Mining Technique: Microsoft Decision Trees
Dimension: Customer
Case Key: Customer.CustomerName
Attributes Used: Gender, Num Of Cars Owned, Num of Childer at home
Mining Model Column Usage: All the above columns except Num of Childeren at Home are Input and Num of Childeren at Home is defined as Predicatble only
Content and Data Type: All the above three columns are defined as discrete and text

Please note that I am new to DM stuff and I might have missed something obvoious. From what I see, when I try to explore the mining model viewer in BI studio, I see only 2 levels and histogram only shows two values, "Missing" and the "", the later having all the values. Ideally, it should have the customer in the root, then split to Gender and then to Num Of Cars Owned. "The num of children at home" should have come fromthe mining prediction.

Thanks for your reply.

|||

When you create the mining model, select the Allow Drillthrough option on the final page of the data mining wizard. When you view the model, right click on the root node and select "Drillthrough". This will show you the data that was used to train the model. It seems that the data that is being used for your model likely isn't the same as was in the book. I am unfamiliar with the book, but if you browse the data and see that maybe Num Children doesn't have the values you expect, it may show you why the model doesn't look like the book's.

You may try e-mailing the book's author as well.

HTH

sql

No comments:

Post a Comment