The evaluation metric for this competition is Accuracy.
For every track in the dataset, submission files should contain two columns: trackId and genre class. The Class can be one of the following genre labels: 'soul and reggae', 'pop', 'punk', 'jazz and blues', 'dance and electronica', 'folk', 'classic pop and rock', and 'metal'.
The file should contain a header (trackId and genres), sorted based on trackId and have the following format:

```
trackID,genre
6732,metal
5415,soul and reggae 
7757,pop
etc.
```

Submissions will be evaluated and ranked by Kaggle for accuracy, against just 30% of the test data, forming the public leaderboard.