“概率输出逻辑回归(科恩alized)"

confuzio

Hello,

using Kernel Logistic Regression in RapidMiner 5.1 I could not figure out by now how to get probability predictions (the estimated probability that this customer will default is e.g. 0.11, not simply: this customer will default). When using "apply model" all I get is 0/1-predictions (RM internally uses threshold p_hat = 0.5?).
As I use the radial kernel I can not simply plug the estimated coefficients in p_hat = exp(Xb)/(1+exp(Xb)).

Can anyone tell my how to get probability predictions? I'd be really grateful!

Here is my process (As I usually work with R I'm not really used to RapidMiner by now):

<运营商激活= " true "类=“检索”兼容ibility="5.1.001" expanded="true" height="60" name="Retrieve" width="90" x="45" y="75">

steffen

Hello confuzio

Here is my dummy process:

<运营商激活= " true "类=“检索”兼容ibility="5.1.001" expanded="true" height="60" name="Retrieve" width="90" x="45" y="210">

which creates confidences aka a rough prob output. What does your process look like ?

In general it is recommended to post the used process whenever possible. Make it easy for us to help you

hope this was helpful,

steffen

confuzio

Ok, thanks for your process. I figured out how to save the confidences.
Just one more question: "which creates confidences aka a rough prob output" -- do you simply mean by logistic regression estimated probabilities? I'm confused about the "rough".

steffen

the answer is: Yes, I mean the estimated probabilities.

Sorry for the confusion

greetings,

steffen

confuzio

Thanks a lot for your help!