0
$\begingroup$

I have a classification problem classifying tumors as benign or malignant. However, I want to go a step further to provide a ranking of these tumors as most malignant to most benign. Are there any good algorithms out there to help with this ranking? Any suggestions?

The features of the dataset are the radius of the tumor, perimeter of the tumor, concavity, smoothness, etc.

$\endgroup$
0
$\begingroup$

Most classification algorithms actually provide continuous scores that are compared to a given threshold to give a binary output. Using that score directly give you a ranking. But unless you give us a specific algorithm, it would be difficult to help you further.

You usually can find that trough the performance metrics. AUC for exemple, has a general definition obtained by varying the above mentionned threshold. However, it has a more natural interpretation (and way faster approximation) when you use a definition based on individual scores.

Some models can even be calibrated in probabilities. That means, under some conditions, the continuous score can be intepreted as a probability to be of a given class.

| 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.