Are there any opportunity to export model to code that a data scientist also can use it?
abanar
New Altair Community Member
Are there any opportunity (in RapidMiner Studio) to export model to code that a data scientist also can use it in R or Python? If it is not possible is RapidMiner make any effort for it?
Tagged:
0
Best Answer
-
Hi,
there are some options.
First, the PMML extension: https://marketplace.rapidminer.com/UpdateServer/faces/product_details.xhtml?productId=rmx_pmml
For models specified in the Predictive Model Markup Language, RM can export them into this standard format.
Other types of models like linear regression give you a formula that is trivial to implement in any programming language.
For decision trees, check out the Tree to Rules operator, it will give you a textual if-then-else based description of the tree splits. This is also easy to implement in any programming language.
Many models are too complex for these approaches, however. There is no simple way to export them to R or Python.
If you have this requirement, you can do the following:
1. Implement your process in RapidMiner
2. Use AutoModel (optional) or your experimenting and validation to find the model you prefer
3. Use the same (or similar) parameters for building the model in an R or Python script, save your model there.
This gives you the flexible and fast experimentation approach that is possible with RapidMiner, but allows you to implement models outside if you need them.
Regards,
Balázs5
Answers
-
Hi,
there are some options.
First, the PMML extension: https://marketplace.rapidminer.com/UpdateServer/faces/product_details.xhtml?productId=rmx_pmml
For models specified in the Predictive Model Markup Language, RM can export them into this standard format.
Other types of models like linear regression give you a formula that is trivial to implement in any programming language.
For decision trees, check out the Tree to Rules operator, it will give you a textual if-then-else based description of the tree splits. This is also easy to implement in any programming language.
Many models are too complex for these approaches, however. There is no simple way to export them to R or Python.
If you have this requirement, you can do the following:
1. Implement your process in RapidMiner
2. Use AutoModel (optional) or your experimenting and validation to find the model you prefer
3. Use the same (or similar) parameters for building the model in an R or Python script, save your model there.
This gives you the flexible and fast experimentation approach that is possible with RapidMiner, but allows you to implement models outside if you need them.
Regards,
Balázs5