AutoML for RapidMiner

cedric_anover

RapidMiner already provides the necessary operators for somewhat building your own AutoML to perform automated supervised learning task. But I also think it would be better if rapidminer can have its own single AutoML operator. This AutoML is similar to TPOT, scikit learn automl, h2o automl, weka automl, etc.

 

Operator Name: RMAutoML

Inputs:

  Data Table

 

Outputs:

  Optimal Model 

  Performance

  Parameter Set

 

Parameters:

  Optimizer: {Bayesian, Grid, Genetic Algo, etc.}

 

MartinLiebig

Dear Cedric,

 

thanks a lot for your feedback. This is definitely a  feature we are aware of.

 

One thing to consider is, that building such a feature the right way is a considerable amount of effort. A big part of the added value of RapidMiner is the combination of simpleness but also completeness of our data science offering. Building a system like you propose in a way that it is easy to use but with good performance opens some challenges. So it's nothing we can code "on the run".

 

But as I said, we are aware of this.

 

Best,

Martin

sgenzer

sgenzer

Nailed it!! https://rapidminer.com/blog/accelerated-modeling-without-black-box/