1
$\begingroup$

I have a dataset of individuals which can be split into two groups: those that have performed an action (group 1) and those that haven't (group 0). For each individual I have a large number of features, some of which can be used to attempt to distinguish between the two groups (including a boolean corresponding to the action I'm interested in).

What I would like to do is to find the subset of group 0 that look the most likely to perform this action, thus moving from group 0 to group 1.

It isn't clear to me what is the best approach to this at all - I started by training a classifier on a subset of my data to distinguish between the two groups. This classifier was then be run over the remainder of the data and I looked at those members of group 0 that were misclassified as being in group 1 (or for which the classifier had a lower probability for their membership in group 0), but this really don't seem to be a sensible approach. Obviously part of what this will be doing is finding a way to reject the 'group 1-like' members of group 0, and these are the individuals I'm really interested in.

Are there any better alternatives? Really what I want is to train a model to find the 'group 1-like' entries in the entire dataset.

Many thanks.

$\endgroup$

1 Answer 1

2
$\begingroup$

You probably moved with your life, but one class classification or P-U learning would be suitable for your problem https://en.wikipedia.org/wiki/One-class_classification

$\endgroup$

Your Answer

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.