GECCO Challenge 2017 Results

Submission Evaluation

For the evaluation of the submitted online-methods, the methods have been executed on a dataset unknown to the participants. The quality of the detector is then calculated using the F1 score, the harmonic mean of precision and recall, which is defined as:

F1 = 2 · (PPV · TPR) / (PPV + TPR)

with

PPV = ∑ TruePositive / ∑ PredictionPositive

and

TPR = ∑ TruePositive / ∑ConditionPositive ,

so that PPV describes the ration between the number of true positive predictions and the number of all positive predictions while TPR the ratio between the number of true positive predictions and the number of actual events. Per definition this score ranges from values from 0 to 1, with 1 being the best achievable score.

Detailed Results

Please click on a row for additional information.

Submission TP FP TN FN F1 Score
Fitore Muharemi 472 636 120670 556 0.44194757
Rishab Dhar, Jagruti Gunjal, Surabhi Jadhav 487 9707 111599 541 0.08679380
Lam Gia Thuan (Crazy method) 242 24392 96914 786 0.01886057
Roman Berneburg, Jan Bickel, Marcus Legendre 234 32662 88644 794 0.01379554
Lam Gia Thuan (Basic method) 297 45920 75386 731 0.01257276

Additional Downloads

Download the GECCO 2017 test data set as RDS file.