Enquiry on model's settings

AizatAlam_129

Hi,

As part of our journal article revision submission, I was asked by the journal article's reviewer to include the specific settings for each algorithm we used in RapidMiner's auto model (we used all except SVM).

We explained that we did not report them for easy understanding. However, we were told that the settings are very sensitive and important to report when machine learning is utilized. For instance, the number of neurons (or, hidden layers) in NN makes a huge difference in the prediction rate. In terms of gradient boosting, which condition (e.g., sci-kit learn, extreme boosting, etc) was utilized? In the case of a random forest, how many trees were set as the range to be selected?

While I look forward to insightful replies from community members here, I'd also welcome and would be more than happy if I could be referred to any resources or publication that can help me with this specific enquiry.

TIA

MartinLiebig

Hi @AizatAlam_129,

there is some parameter optimization going. So the best way to get the chosen parameters is to use the save results button the far lower left of the AM screen. You afterwards have all the details in the repository. Each algorithm has one folder with an object called "Optimal Parameters" which includes the chosen values.

Best,

Martin

AizatAlam_129

@mschmitz, I couldn't find any information relating to the settings of each algorithm under "optimal parameters" at the repository.

MartinLiebig

Hi,

did you store the results of AM first?

BR,

Martin

AizatAlam_129

@mschmitz yes i did