Missing values in classification matrix, flat lift chart when using external data set for testing


  • Hi all,

    I have big data set which is split into train and validation set (in separate tables in db/data source view). What I want to do is to make a decision tree and then test it with external data set. After specifying inputs and external data set I proceed with testing. Results are kinda strange - lift chart is flat line at 0, while classification matrix look like this:

    Counts for 'DT' on [x]
    Predicted Missing (Actual) False (Actual) True (Actual)
    Missing 0 0 0
    False 747707 0 0
    True 30494 0 0

    When I run mining model prediction (probability prediction of attribute x based on prediction join between trained mining model and external test set) I get correct output (for each sample there is calculated probability that attribute x will be true).

    Does anyone have an idea why there are missing values in classification matrix/lift chart, but I can successfully run prediction query?

    ps - I am working with SQL Server 2008R2

    Thanks in advance

    • Edited by Goran F Wednesday, November 27, 2013 9:58 AM
    Wednesday, November 27, 2013 9:39 AM

All replies