I am trying to fit a regression model on a non linear data. The features I have are around 12 and around 800 samples. With the help of PyCaret, i tried to fit the data on to around 22 model, and then selected the best one (Ada Boost) and then tried further to tune it to get better result. However, none of the models gave a positive R2 score, Ada Boost was the least worst performing algorithm. This is the test (red) and predicted test output (green) from the selected algorithm. enter image description here

After trying all various techniques and still not getting a decent result, can we infer that the features are not enough to account for the variation of the target variable ? In other words the provided features don't explain best the target variable.

It may sound silly but am a beginner in Data Sciences, so please dont mind.


1 Answer 1


I don't know much about R2 score, but having it negative all the time seems pretty strange to me. Maybe you should try to use AUC as a metric ( <0.5 classifier is worse than a random classifier, and the closer you're to 1, the best is your algorithm).

If it appears that you still can't find a model giving decent results, the direct conclusion is not that your features are not enough, because there can be plenty of other reasons.

I'd suggest you to try SMOTE, which will create new data based on the ones you already have, and try applying your models again. Sometimes, this is a way to tackle the issue of not having enough data


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.