This website uses Cookies. Click Accept to agree to our website's cookie use as described in our Privacy Policy. Click Preferences to customize your cookie settings.
Accept
Reject
Preferences
cancel
Turn on suggestions
Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.
Showing results for 
Show  only  | Search instead for 
Did you mean: 
Choose Language Hide Translation Bar
0 Kudos

Formalize/Improve the Model Classification Explorere Add-In created by Tarek Zikry

Status: New Submitted by on ‎05-13-2020 09:20 PM
1 Comment (1 New)

I downloaded and ran Tarek's Model Classification Explorer Add-In that he created as a JMP Intern with @KarenC and @mia_stephens. 

 

This is a great add-in for nominal logistic regression for holdout data so that we can apply the logistic equation to the holdout set in its own data file. It's important for a team lead on a modeling project to keep a holdout set under lock-and-key, so being able to grab it (once you think you have the best model of course!) and upload as a new JMP data file and then apply the function from our other file, and explore the performance measures, confusion matrices and ROC curves.

 

HOWEVER, I have had this add-in freeze on me and/or take a very long time (even with small data sets). Yes, it works on some sample data set (Diabetes, Donuts), but in the real world using real data we need to understand how to format it and input it.

 

* COULD YOU FORMALIZE THIS ADD-IN to function in JMP in the next version, as a platform or menu item? And maybe take some of the bugs out? And allow it to work on larger data sets?

Thank you!

See more ideas labeled with:
1 Comment
td-ct
Level I
‎03-23-2021 06:07 PM
‎03-23-2021 06:07 PM

Hello, I also used this add-in recently and I also was worried regarding non-responsiveness auf the add-In, though checking the task manager, it was still working. Thinking about the was it works, one can expect at lest a runtime in the order of O(n^2) without actually checking the code, as it prepares all results in advance. The Donut example has just 100 rows and lasts less than a second. But a dataset with 1000 rows will last around 1 minutes to compute, my dataset with 4000 rows maybe 15 minutes and so on. What is really missing:
- a process bar and a signal to keep JMP alive (the "not responding" condition is just disturbing here!)
- some (hovering) info, what data format is expected:
    - Y: Text, in the form "Feature" and "No Feature"

    - X: a continous number

    - Validation: Optional text, must be "Training" or "Validation"
I do not want to forget to express my gratitude for the authors of this nice add-in. Thank you!