0
$\begingroup$

I have a problem at hand to identify Good/Bad Products using given parameters. The number of parameters are in the order of 5000s and there are multiple values for the parameters. However I do not have a labelled set of data which says these are the products that are good or bad.

For Example, Say the parameters are AX, AY, AZ, B, C, DX, DY, etc. Each of them has a different range. Is decision trees the right approach?

Can classification be applied to this problem?

$\endgroup$
2
  • 2
    $\begingroup$ If good vs. bad is synonymous with common vs. rare, you could anomaly detection, which does not require labels. $\endgroup$
    – Emre
    Commented Mar 15, 2018 at 22:26
  • $\begingroup$ Hi Emre, Thanks for your response, I was looking at anomaly detection. Here the training set is supposed to use only the good examples or is it only the value of the ɛ that can help me identify the bad products? $\endgroup$ Commented Mar 16, 2018 at 5:06

1 Answer 1

2
$\begingroup$

No. Classification requires labelled data. Without labelled data there is no way to solve this. How would you do anything at all, if you don't know which of the products in the training set are good and which are bad? There's no basis for making a decision of any sort.

$\endgroup$
5
  • $\begingroup$ D.W. Thank you for your response. If I can acquire labelled data then what is the approach that will help me solve this problem? $\endgroup$ Commented Mar 14, 2018 at 18:41
  • $\begingroup$ @SumitKumarGhosh, then you should use supervised learning. We'd need more information about the context to give you more detailed suggestions, so if you acquire labelled data, you should ask a new question, and give us details about the features, how you expect they might affect the label, tell us about the problem domain, etc. See also datascience.meta.stackexchange.com/q/2267/8560. $\endgroup$
    – D.W.
    Commented Mar 14, 2018 at 18:46
  • 1
    $\begingroup$ Using CNN'S will ease your work $\endgroup$
    – Aditya
    Commented Mar 16, 2018 at 0:19
  • $\begingroup$ Hi @Aditya, Can you please explain how Convolutional Neural Networks will help in this case? $\endgroup$ Commented Mar 16, 2018 at 5:08
  • 2
    $\begingroup$ Please don't use the comment thread here for discussion or back-and-forth about the problem. Aditya, if you can answer the question that was asked, I encourage you to write a full answer in the 'Your Answer' box below. $\endgroup$
    – D.W.
    Commented Mar 16, 2018 at 5:42

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.