Notations
\[
A =\text{gender}, \hat{Y}=\text{prediction on the income level}, Y=\text{ground-truth income level}.
\]
Final Score
\[
Accuracy := \mathbb{P} [\hat{Y}=Y] 
\]
\[ DDP := \vert \mathbb{P}[\hat{Y}=1|A=0] - \mathbb{P}[\hat{Y}=1|A=1]\vert
\]
\[ Th := 0.1
\]
\[
\text{ If } DDP > Th, Bias :=  7^{DDP-Th} - 1,  \text{ else, } Bias := 0.
\]
\[ Score := Accuracy - Bias  \]
The leaderboard will only show your prediction's Accuracy on 50% randomly selected testing points. Your final score will be estimated offline after the submission is closed.
Submission Format
Submission files should contain two columns: id and income. id should go from 0 to 8996. income should be your prediction on an individual's income level (i.e. 0 or 1).
For an illustration, please check out
test_sample1.csv
in the
Data
section.