Auto Documentation

Nirr3

One of the most time consuming (and boring) parts of the data science journey is documenting the model. RapidMiner has a lot of nice visual components to do exploratory data analysis of as you build your model. I often time find that I either have t I take screenshots of the visuals and put them in some sort of a word document. 

It would be really nice to simply be able to do export a process flow with some standardized visuals into say word so I can simply edit and modify. I think if this could get me 80% of the way there, I’d save a ton of time. 

I think this could be something that follows a CRISP-DM documentation style even better.

sgenzer

Hi @Nirr3 - for this have you tried the "Projects" extension? It sets up a nice template for organizing your projects.

Scott

Nirr3

@sgenzer

Yeah. I found the extension super useful to keep my work organized. 

For the auto documentation it’s more about the last leg of the data science work flow. When the model goes into reviews etc. before deployment  we need it to have some sort of documentation. To me the standard is the CRISP- DM methodology as pictured below. 
CRISP- DM has a bunch of reports like data quality, assumptions models make etc that shouldn’t be too bad to incorporate into a word document from
RapidMiner. It doesn’t have to be exactly the way I would write it, just that it covers a lot of the basics.

H20.ai has a similar feature where they generate a lot of the documentation out of the gate based on the processes you run. 

YouTube video below
https://youtu.be/W1ALky-wHBU

I personally think this will help speed up modeling efforts much more.

sgenzer

yes I get it and it all makes good sense. Pushing to feature requests with your other ideas this morning!