The evaluation metric for this competition is F1-Score. It measures accuracy using the statistics Precision and Recall. Precision is the ratio of true positives (tp) to all predicted positive samples (tp + fp). Recall is the ratio of true positives to all positive samples (tp + fn). The F1-Score is given by:

F(f) = \frac{Precision \cdot Recall}{Precision + Recall}]

Before the submission deadline , you can only view the F1-Score evaluated on 50% of the dataset on the leaderboard. After the deadline, we will reveal your model F1-Score evaluated on the entire dataset to rate your performance. However, you do not need to worry that your score would be affected by 1 or 2 percent of F1-Score loss; we will mark based on the baselines your model defeats.

Submission Format
Your submission file should contain two columns: Id and Probability. Each test author node pair is allocated a unique id, and you need to use this id as your prediction index. For example, the first row 3280,50982 means a test node pair with author id 3280 and paper id 50982, its index is 0. If you predict there exists an edge between 3280 and 50982 with prediction 1, the value of the corresponding row of your uploaded file is 0,1.

The file should contain a header and have the following format:

Index,Predicted
0,1
1,0
Detailed description can be found in ‘data’ page.