2
$\begingroup$

I am performing anomaly detection using K-Means. I am working with only one column, plotting those values and then within this column I am adding some anomalies.

enter image description here

My question is

if it is possible to perform confusion matrix for this ? If so what is the approach ? Asking this since I have seen that confusion matrix needs y axis and x axis

$\endgroup$
1
$\begingroup$

Confusion matrix needs groun truth values and predicted values.

You have _____, so you need other part for this multiclass confusion matrix.

| improve this answer | |
$\endgroup$
1
$\begingroup$

No, a confusion matrix doesn't make sense here. While you are assigning inputs to clusters, you do not know which cluster is 'correct' for each input. That is, it is not a supervised multi-class classification problem. You do not even know what each cluster "means".

All you can measure here are metrics like intra-cluster distance: how far on average are points from their assigned cluster? Lower is better. Metrics like silhouette score are probably even better: https://en.wikipedia.org/wiki/Silhouette_(clustering)

| improve this answer | |
$\endgroup$

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.