The evaluation metric for this competition is Accuracy. I don't think it necessary to provide a formula for the same.
Submission Format
You need to submit classes of 159 tuples in the test dataset. The submission file must contain exactly two attributes: 'ID' of the tuples and its corresponding 'Class' label.

The submission file should contain a header and have the following format:
ID,Class
371,1
372,3
373,0
374,5
etc.
For your reference a sample_submission.csv has been uploaded in the Data tab that will clarify any doubt regarding submission format.
You are given 3 days with 7 submissions on each day.

Your submissions will be evaluated on 70% of the test dataset to give your results on the public leaderboard. At the end of the competition, your selected submissions (you can select maximum 2 submissions) will be evaluated on rest 30% of the test dataset and will be shown on the private leaderboard. However, your final marks will be based on your performance on 100% data.
Note:
AFTER the competition ends, you need to upload the jupyter notebook that gave you your best result (corresponding to the public leaderboard) as a kernel on Kaggle. The result obtained on running your uploaded kernel must match your score on the private leaderboard. In case of mismatch, no marks will be awarded.