Our JRS'2012 Data Mining Contest has finished!
The competition attracted a total of 396 teams with 533 registered members. Among them there has been 126 active teams who submitted at leas one solution to the leaderboard. A total number of submissions was 5964. The participating teams were from 50 different countries across six continents. From the active teams, 41 provided us a brief report describing their approach.
Thank you very much for your hard work!
You were able to significantly improve the baseline result. The improvement in the F1score exceeded 125% (the best team obtained a result of 0.53579, while the lowest baseline was 0.23721).
We have also invited nine teams to contribute an extended version of their report to a special session at the JRS'2012 conference, which will be devoted to our contest.
The true classification of the test data is now available! (here)
Other meta-data related to the competition are available below. They include true column names, label names, as well as PMID identifiers of documents corresponding to the training and test cases.
If you are planning to use our data in your research, please give us credit by referring to the contest in your related publications. A bibtex entry of a paper describing our competition is available here.