## Evaluation Metric: Macro-Averaged F1-Score

Submissions are scored using the **macro-averaged F1-score**, which calculates the F1-score independently for each sentiment class (positive, negative, neutral) and then takes the unweighted average of these scores.

This metric ensures that all classes are treated equally, regardless of their frequency in the dataset. It encourages balanced performance across all sentiment categories, preventing models from favoring the majority class.

The F1-score is the harmonic mean of precision and recall, defined as:

`F1 = 2 × (Precision × Recall) / (Precision + Recall)`

where:

- **Precision** measures how many of the predicted positive examples are actually positive.
- **Recall** measures how many of the actual positive examples were correctly predicted.

Using macro-averaged F1-score helps evaluate the model's ability to correctly classify all sentiment classes fairly.