PAIR-code / what-if-tool

Source code/webpage/demos for the What-If Tool
https://pair-code.github.io/what-if-tool
Apache License 2.0
918 stars 170 forks source link

Cant use fairness solution for regression so In that case How to mitigate the bias for the regression model? #226

Open Akankshaw opened 1 year ago

jameswex commented 1 year ago

I'd recommend looking at your model's performance sliced by different subgroups of the dataset to see on which groups the model is performance best/worst at. And then trying to find more training data in those areas of concern to get a better trained model.

In general, fairness evaluation for regression models is not as simple as binary classification, as many of the techniques and analysis depend on having a positive and negative class to investigate. Another approach is to turn your regression problem into a classification problem and using the binary classification fairness tools to understand more about your model in that context.