Auto Model Performance

koichi

In the over view of Auto Model, there is Error and Standard Deviation. 

The description of performance in the information is;

"Performance: lists the model's prediction accuracy and other performance criteria, depending on the type of classification problem. The performance is calculated on a 40% hold out set which has not been used for any of the performed model optimizations. This hold-out set is then used as input for a multi-hold-out-set validation where we calculate the performance for 7 disjoint subsets. The largest and the highest performance are removed and the average of the remaining 5 performances is reported here. Although this validation is not as thorough as a full cross-validation, this approach strikes a good balance between runtime and model validation quality."

I think it is the  standard deviation of the remaining 5 performances. Am I correct?

Regards,
Koichi

MartinLiebig

Hi,

indeed. The way how we calculate it is a bit sophisticated. But you are correct with your statement.

Best,

Martin

koichi

Martin,
Thanks a lot．
Kochi