2022 MEBDI ML Competition
The second annual MEBDI Machine Learning Competition was held in 2022.
The objective of the 2022 ML Competition was to devise a ML classifier algorithm with the best out-of-sample performance for predicting the unemployment status of workers one year ahead. The sample was drawn from the US Current Population Surveys, covering 2008 to 2015. Further details and competition rules can be found here: 2022 MEBDI ML Competition Details
Submissions were evaluated by a two-judge external faculty panel (who are also former alums)—Professors Amy Handlan from Brown University and Carter Braxton from UW-Madison. The panel reviewed the submissions to verify the results and compliance with the competition rules.
This year the second place team showed an excellent performance and lost to the winner after a tie-breaker. So, in addition to the Grand Prize awarded to the winner, the MEBDI award committee decided to award a $1,000 prize to the second place team.
We are grateful to Litigation Analytics, Inc., for funding this year’s Grand Prize.
Confusion Matrix of Baseline Logistic Algorithm. Fractions classified in parentheses.
Confusion Matrix of Winning Algorithm. Relative to the baseline, the fraction of unemployed classified correctly rises from 3.7% to 71%, while fraction of employed classified correctly falls from 99.8% to 79.9%.
Winners
Grand Prize, $5,000
Awarded to:
Robert Winslow
Read the Executive Summary of Robert’s
submission:
Read the Judging Committee’s Summary:
SECOND PLACE PRIZE: $1000
Awarded to the team of:
Thomas Hasenzagl and Nesile Ozder.