Ways to deploy ML model written in PMML

Hi there!

Recently I had to conduct research on ways of deploying a pmml model. I spent some time finding different options and now I decided to make my findings available for publicity to save your time.

Let’s go!

IBM Watson

Paid: yes.

There is an option to use it for free with limitations: 50 evaluations per hour or 5000 a month.

REST: yes.

Ability to put your pmml file on web-interface.

It logs web-requests.

Instructions:

https://www.ibm.com/support/pages/how-do-you-deploy-r-pmml-model-watson-machine-learning-service-watson-studio
https://www.youtube.com/watch?v=JuNQJwfP2uI

Clouderizer

Paid: yes.

REST: yes.

Ability to put your pmml file on web-interface.

It logs web-requests.

Ability to deploy models on AWS, Google Cloud Platform, your PC under Ubuntu/Mac OS.

Instructions

https://medium.com/@prakash_31206/deploy-your-pmml-models-to-production-in-under-two-minutes-e7a3029a1149
https://www.youtube.com/watch?v=52dec-IQLVQ&feature=emb_logo

Domino Data Lab

Paid: yes.

REST: yes.

Ability to put your pmml file on web-interface.

It logs web-requests.

There is a try-period available.

Instructions

https://docs.dominodatalab.com/en/4.3.3/reference/publish/models/Model_publishing_overview.html
https://docs.dominodatalab.com/en/4.3.3/reference/publish/models/Model_deployment_configuration.html

Plumber

Paid: free.

It’s an open source package in R that provides easy ways to create a RESTful API for programs written in R. If you are familiar with Python or Ruby, it is an equivalent of Flask and Sinatra.

REST: yes.

Technology stack: R

Instructions

https://www.knowru.com/blog/how-create-restful-api-for-machine-learning-credit-model-in-r/

https://www.rplumber.io/

Zementis by Software AG

https://www.softwareag.com/ru/products/az/zementis/default.html

Paid: yes.

REST: yes.

Ability to put your pmml file on web-interface.

Instructions

https://support.zementis.com/hc/en-us/ — help desk

https://www.youtube.com/watch?v=ovFEBusKdNs&t=338s

https://www.youtube.com/watch?v=3UBiSVsJwZU

Openscoring

https://openscoring.io/

Paid: free(GNU Affero General Public License v3.0).

Openscoring is an open-source project providing us with a bunch of libraries around PMML.

Technology stack: Java/Python/R

Instructions

Openscoring.io blog

JPMML-Evaluator: Preparing arguments for evaluation

Deploying LightGBM models on Java/JVM platform

Detailed documentation with Java examples

Syncfusion PMML

https://www.nuget.org/packages/Syncfusion.PMML.AspNet/

Syncfusion PMML is a .NET library for predicting results based on predictive models written in Predictive Model Markup Language (PMML).

Different versions are available on NuGet.

Paid: yes.

Technology stack: .NET