## Evaluation

The evaluation metric for this competition is categorization accuracy. It is calculated by the fraction of predictions our model got right. Formally, accuracy has the following definition:

![enter image description here](https://tutorialspoint.dev/image/Confusion_Matrix2_2.png)


TP: True Positive
TN: True Negative
FP: False Positive
FN: False Negative

## Submission Format
You should submit a file containing the predictions for the `test_data.csv` images. The submission file should contain two columns: ID and Emotion
For more details, see the [example_submission.csv](https://www.kaggle.com/c/expertclass2/data?select=example_submission.csv) file.

## Leaderboard

The `test_data.csv` is pre-divided into Part A and Part B to generate a public and a private leaderboard. Part A will be used to generate the accuracy of the model (displayed public) and Part B to check possible overfittings. Due to that, the results may vary if the solution is not generalizing well, but you will be notified if it's happened.  
  
Each part will have a weight of 50% in the result. The model must have good accuracy and the ability to predict unknown data.

Final score = (public leaderboard score * 50%) + (private leadership score * 50%)