Handling Imbalanced Data

Would like explore options/ solutions to cases where we have imbalanced data set. For example we have 99 positive cases and only 1 negative cases or vice versa.

Though we can find such articles across Medium, any personal experience and strategies are welcome.

This looks more like a problem of finding/identifying outliers in the dataset.

If it really needs to be considered as 2 classes, one possibility is to assign a high weight to negative class (for e.g. 99) and a small weight to positive class (lets say 1) . Better to make the weight assigned a hyperparameter (and then play around with different values to see what works best)

I fear if just one out of 100 cases is negative, model will underfit.

Thanks Ishvinder. Was thinking of generating more artificial samples