1
$\begingroup$

I've created a binary classifier using K Mean, which predicts fraud and legitimate accounts, 0 and 1. This uses two features, let's say, A and B.

enter image description here

Now, I want to use other features like C and D, to predict fraud and legitimate accounts. Can I use output class predicted in Kmeans, and use it as a label for logistics regression using features C and D?

Sorry, if this sounds like dummy questions, I'm just starting into Data Science, so please let me know if I missed something fundamental.

$\endgroup$
3
  • $\begingroup$ Don't use k-means for prediction problems. If you run it twice, you'll get different results... $\endgroup$ Mar 6, 2019 at 7:06
  • $\begingroup$ What's best algorithm for predicting category for unlablled data? $\endgroup$
    – Mukul Jain
    Mar 6, 2019 at 7:41
  • $\begingroup$ No clustering will predict a real "category". Categories are a supervised concept. They can enumerate groups though. $\endgroup$ Mar 6, 2019 at 18:56

1 Answer 1

1
$\begingroup$

It is strange to use k-means in addition to logistic regression. Usually k-means is reserved for unsupervised learning problems, this is when you do not have labelled data. Unsupervised learning algorithms are not as powerful and it seems here you have labelled data, thus you should stick to supervised learning techniques.

If you have additional features for your instances then you should include them in your logistic regression.

$\endgroup$
1
  • 1
    $\begingroup$ Thanks for the answer. So to clarify, I don't have labels in starting, I predicted them using k-means. Now, I've target field, so can I run logistics regression using different features (other than what I used for k-means) and use k-means predicted output as label? $\endgroup$
    – Mukul Jain
    Mar 6, 2019 at 4:47

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.