7
$\begingroup$

Can I use the probability score generated from a Machine Learning model as a feature in another model? For example, say we have a model which generates the probability of an ad being bad. Lets call it badness_score. I am working on building another model which will predict the probability of an advertiser being a fraudster. Can I use badness_score as a feature while building this model ? Are there any caveats to this approach?

Improve this question
$\endgroup$
4
  • $\begingroup$ You can use it as a feature, sure. However I'd imagine the utility of this feature being strongly connected to the quality of the first model - if it captures the underlying pattern poorly, it's probably not beneficial to the second model. $\endgroup$
    – deemel
    Commented Apr 24, 2019 at 13:57
  • $\begingroup$ I am facing a similar problem. I have a model that predicts the sentiment of an input sentence (bad or good). I want to use the output from this model as an input to another one that predicts if the content of the sentence indicates sexual harassment. However, since the input features are text sequences, I don't know how to use the binary output from the first model into the second model. $\endgroup$
    – Alexandre Henrique
    Commented Mar 10, 2022 at 13:48
  • $\begingroup$ Welcome to the site. Was this intended as an answer to the OP's question, a comment requesting clarification from the OP or one of the answerers, or a new question of your own? Please only use the "Your Answer" field to provide answers to the original question. You will be able to comment anywhere when your reputation is >50. If you have a new question, click the blue ASK QUESTION at the top of the page & ask it there, then we can help you properly. Since you're new here, you may want to take our tour, which has information for new users. $\endgroup$
    – kjetil b halvorsen ♦
    Commented Mar 10, 2022 at 15:08
  • 1
    $\begingroup$ Such a "model chain" is implemented in sklearn, fwiw: scikit-learn.org/stable/modules/multiclass.html#classifierchain $\endgroup$
    – Ben Reiniger
    Commented Mar 10, 2022 at 15:22

1 Answer 1

Reset to default
6
$\begingroup$

It's perfectly OK, and also widely used. Different models can explain different perspectives of data and stacking them in front of each other, and using outputs/predictions produced by previous layers enables even moderately simple final layer algorithms to perform much better compared to on their own, because they use the cumulative knowledge learned via other algorithms. This is somewhat analogous to adding layers to neural networks.

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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