3
$\begingroup$

I doubt better is the correct adjective, apologies for that. What I mean is this: I have a set of files (1200~) each paired with a scatterplot image. I need to find a way to classify which data files will have scatter plots that a person would classify as "clearly separated data"(again, not the right words), and which ones as "not clearly separated". For example: The data in this scatterplot is clearly separated,

The data in this scatter plot is separated more or less clearly

The data in this scatter plot is not clearly separated

The first and second scatter plots are examples of data that would end classified as "good" or "clear", while the third one would be classified as "unclear". Are there metrics or distances that could be applied on the data to predict wheter the plots generated by that file would be of either the first kind or the second?

Thanks for taking the time to read this. I'm new to this, and english is not my first language.

New contributor
Gaverfraxz is a new contributor to this site. Take care in asking for clarification, commenting, and answering. Check out our Code of Conduct.
$\endgroup$
  • $\begingroup$ Can you label manually about 100 of them? And then train a model to classify data, and use another small set to validate. A small Random Forest could do the job well. $\endgroup$ – Leevo Feb 14 at 9:07
1
$\begingroup$

You could:

  1. Label manually few of them (say 100-150), then train a simple model to classify data. A small Random Forest could do the job well.

  2. Train a super basic model on the each dataset used to produce each scatterplot. Something like a Linear Classifier. If the Classifier doesn't make mistakes, you have "clearly separated data", if it makes mistakes then it's likely to be the opposite.

$\endgroup$

Your Answer

Gaverfraxz is a new contributor. Be nice, and check out our Code of Conduct.

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.