The data I am working on has some really large price values and some really small values. What I did was first perform feature bagging on the data and got them labelled to (0,1) and then did Clustering on the data along with the labels found in the previous task.

Is this a right way to go?

  • 1
    $\begingroup$ Since clustering algorithms are very sensitive to scaling and tend to have issues with non-continuous attributes and attributes of different types I'd assume you are just making things worse. $\endgroup$ – Has QUIT--Anony-Mousse May 23 '19 at 5:07
  • $\begingroup$ "Is this a right way to go?" Probably depends on your data and on what you want to achieve. There may be data where this is exactly the right thing. Maybe you could specify a bit more, how the data is distributed and also specify what your goals are? $\endgroup$ – Trilarion May 23 '19 at 14:41

What you could try is to make a transformation of the price data, i.e. log(price). By doing so, very large values become smaller. The data is more "harmonic" after the transformation.

|improve this answer|||||

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.