Hi there,

I want to perform different Machine Learning approaches for survival analysis in cancer.

I have a database of 330 patients. I have some unique evaluated features so I can't use online database to look for other cohorts of patients to use as train or test.

How can I manage this? Should I skip the training cohort? Or can I use my test cohort as training and then generate a new random cohort? How? Is this possible? Is there any protocol or recommendation to follow?

Or should I split my database in 150 and 180 patients for example?

Thanks in advance,

Carlo