[Please Help Needed] ideas on how to handle this problem

[Sere995]

Hi there!

I have a dataset formed by 3 columns : userId, Class and a matrix.
I have 5 possible classes.
I would like to try to handle it with the techniques of multi-instance learning but I couldn’t find anything online… Do you know what plug-ins I coud use or do you have an example of knime with multi-instance learning?

Or do you have any suggestion on how to handle such a problem? Any input would be greatly appreciated, I’m really desperate… thank you so much in advance

[Snowy]

I’m not familiar with MIL, but my best guess is you can use KNIME to do your data preprocessing, then use a package such as Python’s MIL (mil · PyPI) or R’s MILR package (CRAN - Package milr) to actually do the model training.

[Sere995]

Thank you so much for your answer! I’m trying to follow your advice… so I should install one of those two packages in knime, for example using the Python Edit Variable node?

[Snowy]

So I have only used R within KNIME, so I can only provide context in that regard… Typically, after you have your data set ready, I use the “Table to R” node. Within that node, I install my packages (e.g. install.package(“milr”) — highlight and click Eval Selection) and type my script…

If you’re not familiar with R, I’d probably suggest viewing a few examples on the KNIME Hub: table to r – KNIME Hub then view the Vignettes examples: milr: Multiple-Instance Logistic Regression with Lasso Penalty or the reference manual: https://cran.r-project.org/web/packages/milr/milr.pdf

Of course give it your best and we’ll be here to help

[ipazin]

Hello @Sere995,

and welcome to KNIME Community!

To use Python or R nodes within KNIME you need to perform some integration steps. Here is link to KNIME documentation page where you can find Python and R guides which help you do it:
https://docs.knime.com/?category=integrations&version=latest

(Additionally please don’t open multiple topics for same question.)

Br,
Ivan