I'm trying to artificially create a dataset for pure educative reasons but I want it to be based in one particular dataset, the problem is that this original dataset don't make good predictions even with the most powerful methods (with the predictive variables you can't guess the target because they don't have a strong relationship)

My approach to this problem will be using an oversampling for imbalanced data method, as it usually creates artificially new observations intepoling data from the predictible variables..

I also have remember this. There are methods for filling NAN values that interpolates them with different methods throught the rest of the dataset. Maybe filling with NAN values random observations of the dataset and then using this method could work.

Or maybe try out synthetic data, an open source example would be ydata-synthetic, eventhought I dont know how to apply it.

Other option I think it could be first of all training the model. Then, whatever data points are not being predicted well, you can throw those out. You could even repeat the process a couple times, only throwing out a small percentage each time.



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.