when I look into the R package pmml, I found that, it is possible to directly export nnet model into PMML format.

But actually I am using neuralnet package since it has more features and flexibility.

The problem is:

How can I export neuralnet model to PMML format, I checked pmml package, it doesn't work.

Any other options like transform neuralnet model to nnet model, or some other PMML packages to use? thanks


You already mentioned the alternatives...

  • convert the neuralnet object to a nnet class object
  • implement pmml.neuralnet()

Both pmml, nnet and neuralnet packages are easy to read. You can find the source code over cran.

I think implementing pmml.neuralnet() might be the cleanest. You can just copy-paste the current pmml.nnet() method, call it pmml.neturalnet() and make the necessary changes. Should not take more than a day's work.

ps: It would be cool if you'd sent your function to neuralnet authors afterwards of course, and link it from here. :)

| improve this answer | |
  • $\begingroup$ I already contacted the author Frauke Guthener who wrote the neuralnet, but he seems to have no interest on transform nn model to pmml model. and I started to work on the suggestions you gave. and since I am full-day employee and have a girlfriend who likes to travel a lot... I can't foresee the date i will finish that thing. I will paste a link for that work when I finish it. and thanks @Ricardo Cruz $\endgroup$ – cinqS Dec 26 '16 at 2:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.