"Feedback on RapidMiner vs. Python comparison"

mlg1988

Hello everyone,

I made a comparison between RapidMiner and Python for Data Science.
I share an image with this post. I started using RapidMiner only recently, I was using mostly Python so far,
so any feedback is welcomed. Thank you.

rfuentealba

Hello, @mlg1988

Some precisions:

For item: "1 data science project", you can use building blocks, subprocesses or other methods to make RapidMiner do one to many mappings. It is possible to reuse code, or even models if you want to create a model and use the Store operator to save it and use it elsewhere. You already mention building blocks later in your document.

For item: "get help as you work", as @varunm1 said, we are usually around, asking stuff and talking with each other about the inner things of data science (also: mocking each other on Twitter too, all friendly and amicable), so yes, you can count on the RapidMiner Community. Hint: even if you are rude with us you will find some people who will answer your questions in a nicely way. My best advice is: don't push it, but it has happened lots of times. People here are so kind that they make me look like a helpless unmannered Neanderthal.

For item: "Project sharing", if you use a Java-supported Desktop Environment (Windows, macOS, or a FreeDesktop-compatible UNIX Desktop Environment/Window Manager), you can share your entire RapidMiner repository together with data and project by clicking with the secondary button and selecting "Open in file browser". I normally use Git there, so that I can sync my projects on my many computers, for example.

Other than that, thank you, it looks good.

Hope this helps,

Rodrigo.

varunm1

Hello @mlg1988

The comparison looks great. I think you can add this community which is very helpful in "get help as you work" as this is a dedicated community that is quick in answering user questions and concerns.

Also, the "knowledge level needed". In my view python needs medium coding experience whereas RM needs a conceptual understanding of how things work as rest is drag and drop operators. This can also be stated as ease of use.

rfuentealba

Hello, @mlg1988

Some precisions:

For item: "1 data science project", you can use building blocks, subprocesses or other methods to make RapidMiner do one to many mappings. It is possible to reuse code, or even models if you want to create a model and use the Store operator to save it and use it elsewhere. You already mention building blocks later in your document.

For item: "get help as you work", as @varunm1 said, we are usually around, asking stuff and talking with each other about the inner things of data science (also: mocking each other on Twitter too, all friendly and amicable), so yes, you can count on the RapidMiner Community. Hint: even if you are rude with us you will find some people who will answer your questions in a nicely way. My best advice is: don't push it, but it has happened lots of times. People here are so kind that they make me look like a helpless unmannered Neanderthal.

For item: "Project sharing", if you use a Java-supported Desktop Environment (Windows, macOS, or a FreeDesktop-compatible UNIX Desktop Environment/Window Manager), you can share your entire RapidMiner repository together with data and project by clicking with the secondary button and selecting "Open in file browser". I normally use Git there, so that I can sync my projects on my many computers, for example.

Other than that, thank you, it looks good.

Hope this helps,

Rodrigo.

IngoRM

Hi @mlg1988

Fantastic comparison indeed!  In addition to what the others have said already, I would like to add a couple more points:

• It looks like you focused mainly on RapidMiner Studio and within that the process designer.  And I would agree that this certainly is the key part of it so that makes a lot of sense.  In case you want to extend this a bit, there are some other elements of the platform which I personally think may be relevant to your comparison:
  (1) RapidMiner Server - relevant for "project sharing" (shared repositories) and collaboration (allows multiple users to work on the same project including access rights management etc.);
  (2) Auto Model - relevant for "get help as you work", let's users quickly do some prototyping and allows model building even without process design (but full processes are created in the background)
  
• On "programmability": you may want to add that you can also integrate scripts written in Python, R, Groovy, or Java in addition to the extension writing.

Sorry if some of the above sound a bit like marketing, but, you know, I may be a bit biased My main point though is that, as a platform, there a couple of other elements which I personally think may be important to consider.

Hope that helps and I think it would be awesome if you could share the final result with us.  I think many people will find it very helpful to understand when to use what.

Best,
Ingo