Skip to main content

Timeline for How to correctly evaluate a model?

Current License: CC BY-SA 4.0

11 events
when toggle format what by license comment
Aug 12, 2022 at 19:38 comment added dipetkov It matters where the shuffle=True is added. Anyway, I think the best use of your time will be to learn enough scikit-learn. And the best way to achieve this is to read the docs and do the tutorials.
Aug 12, 2022 at 19:33 comment added Z47 I think I have shuffle=True in the code I posted.
Aug 12, 2022 at 19:27 history edited dipetkov CC BY-SA 4.0
added 112 characters in body
Aug 12, 2022 at 19:25 comment added dipetkov Oversight on my part: I didn't show you how to shuffle the data, as scikit-learn KFold doesn't shuffle by default. You have to know how to turn it on. The most efficient way to learn this is reading the documentation.
Aug 12, 2022 at 1:00 comment added Z47 Thank you. I've figured that out that's why I've deleted my comment. The thing is now my R^2 is -216 and using kfold it was 0.42!
Aug 12, 2022 at 0:53 comment added dipetkov As the comment says, this is the average R-squared across the 10 folds. You can also get the scores for each fold. You will need to actually read the documentation to learn how to use these tools.
Aug 12, 2022 at 0:20 history edited dipetkov CC BY-SA 4.0
added 845 characters in body
Aug 11, 2022 at 22:47 comment added Z47 let's assume GridSearchCV is not an option, may I know which approach you would go with and why?
Aug 11, 2022 at 22:45 comment added dipetkov I think GridSearchCV is a better option than either approach 1 or 2, so that's why I wrote the answer.
Aug 11, 2022 at 22:39 comment added Z47 Thank you. I've used RandomizedSearchCV and GridSearchCV with CV in each but I haven't included this part of my code in the question. I was hoping for a more of a straightforward answer to my question. Like which approach 1 or 2 or both are acceptable?
Aug 11, 2022 at 22:24 history answered dipetkov CC BY-SA 4.0