Cookies help us deliver our services. By using our services, you agree to our use of cookies. Learn more

crowdAI is shutting down - please read our blog post for more information

IEEE Investment Ranking Challenge

Explore methodology to identify assets with extreme positive or negative returns.


different metric in overview and github example

Posted by cyga over 2 years ago

At overview https://www.crowdai.org/challenges/ieee-investment-ranking-challenge you have: n/(n+i) in DCG coefficient. Meanwhile looks like you have n/(n-i) in github metric https://www.crowdai.org/challenges/ieee-investment-ranking-challenge I checked the data inside metric function and I tried to train it with “+” instead - it gave me worse result.

Please clarify: * it has to be minus in overview * or it has to be plus in github (and why then result trained with plus is evaluated worse on LB?)


Posted by benharlander  over 2 years ago |  Quote

@cyga Thanks for the heads up! I assume you were referring to the DCG calculation in the random forest starter kit. The DCG in the overview is the correct formulation. We will update the starter kit to reflect this.