This example uses AUC values to check the performance of the model used in Example 1.
Input
The input table is roctable, as in ROC Example 1: Show Only ROC Values.
SQL Call
Because this call specifies AUC ('true') and omits the ROCValues argument, the ROCValues argument has the value 'false'.
SELECT * FROM ROC ( ON roctable AS InputTable OUT TABLE OutputTable (rocoutput21) USING ModelIDColumn ('model') ProbabilityColumn ('probability') ObsColumn ('obs') PositiveClass ('p') NumThreshold (100) AUC ('true') ) AS dt;
Output
info |
---|
ROC complete. |
model | auc | gini |
---|---|---|
1 | 0.375 | |
2 | 0.375 | |
3 | 0.375 | |
4 | 0.375 |
The AUC values are less than 0.5, which means that the model performs worse than random guessing, and you must change it before you use it for prediction. Compare these AUC values to the AUC value in the ROC Example 3: Show AUC and Gini Values output.