Frequently Asked Questions
KDD Cup 2008: Breast cancer
Announcements
The evaluation function get_ROC_KDD.m has been changed slightly on Apr 24, 2008 to fix a minor bug. Please download and use the new version (requires login). We would like to thank Will Dwinnell for identifying the bug.
Another bug was found in get_ROC_KDD.m. The file has been changed on May 27, 2008 to fix the bug. Please download and use the new version. We would like to thank Peter Gehler and Sebastian Nowozin for identifying the bug.
get_ROC_KDD.m was updated on June 19, 2008 to fix a bug. Please download and use the new version. We would like to thank Peter Gehler and Sebastian Nowozin for identifying the bug.
get_ROC_KDD.m was updated on June 23, 2008 to fix a minor bug. Please download and use the new version. We would like to thank Sinno Jialin Pan for identifying the bug.
Frequently Asked Questions
How do I format the results files?
Challenge 1 - The results file should be a list of scores (one per line) for each candidate which corresponds to the candidates exactly as given in the test file. The scores can be in the range -infinity, +infinity, or [0,1]. As long as the low score indicates low likelihood of malignancy and high score indicates high likelihood of malignancy, and as long as the get_ROC_KDD.m file works with your scores, our evaluation script will work. You can refer to pred_scores.txt to see example of the format.
Challenge 2 - The results file should be a list of patient_IDs and predicted label (-1 for benign and 1 for malignant patient), one per line, patient_ID and prediction separated by comma. Please see pred_per_patient_labels.txt for a sample file.
Note that the example files are provided as an example for format. The scores were generated completely at random.
How do I submit the results files?
The submission page is coming up early the week of June 16.