enter image description hereArgument 1
Your model classifies 80% of group A as “unlikely” but 80% of group B as “likely.” This data scientist complains that the model is treating the two groups unfairly in the sense that it is generating vastly different predictions across the two groups.
Argument 2
Regardless of group membership, if we predict “unlikely” you have a 20% chance of action, and if we predict “likely” you have a 60% chance of action. This data scientist insists that the model is “accurate” in the sense that its predictions seem to mean the same things no matter which group you belong to.
Argument 3
40/125 = 32% of group B were falsely labeled “likely,” whereas only 10/125 = 8% of group A were falsely labeled “likely.” This data scientist (who considers a “likely” prediction to be a bad thing) insists that the model unfairly stigmatizes group B.
Argument 4
20/125 = 16% of group A were falsely labeled “unlikely,” whereas only 5/125 = 4% of group B were falsely labeled “unlikely.” This data scientist (who considers an “unlikely” prediction to be a bad thing) insists that the model unfairly stigmatizes group A
Explain (Data science)
Xuân Huy Nguyễn is a new contributor to this site. Take care in asking for clarification, commenting, and answering.
Check out our Code of Conduct.
1