I have a labeled dataset and I created a duplicate of this dataset and removed the labels and applied K-means clustering with k= the number of labels in the original data set I want to compute similarities between labels to test the efficiency of the algorithm any ideas?

  • $\begingroup$ One thing you can do is , you have data A - which contains labels and data associated with it, you have another data B, with labels(clusters) and data associted with it. You can look at how your features are distributed and understand the similarities. $\endgroup$ Mar 18, 2021 at 15:14
  • $\begingroup$ I have 26 feature and that is a lot to compare should i do some dimensionality reduction to ease the processor do I have to do it with 26 features $\endgroup$ Mar 18, 2021 at 15:31
  • $\begingroup$ I think you can do chi-square test or calculate cohen kappa score to validate your model efficiency. $\endgroup$
    – Ram
    Mar 18, 2021 at 19:10
  • $\begingroup$ i want to calculate based on the features of each class ie:comapre each created class with each original class based on features not labels themselves and see the similiarties $\endgroup$ Mar 19, 2021 at 22:06


Your Answer

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