I am developing an ML model for classification using tabular data. It has 5 classes right now and new classes are expected to be continuously added. (Already have a new one leading to an imbalanced dataset). Data is noisy but it has no missing values.

I have a single ML model based on an ensemble of Lightgbm, Xgboost, and TabNet. I have tuned the hyperparameters and believe that not much could be done about it anymore.

However, would it be better to have a separate one vs all model for each class? Or a cascade model considering the imbalance in the dataset?

What are the pros and cons of the two different approaches?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.