Submissions will be evaluated using a custom two-stage metric based on the absolute percentage error between predicted and actual values. The custom score is computed from a scaled [MAPE](https://en.wikipedia.org/wiki/Mean_absolute_percentage_error) with two stages.

**First stage:**

If over 30% of the submitted samples have absolute percentage errors exceeding 100%, a score of 0 is immediately given.

![score_0.png](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F28096579%2F573450f7364d161293c2ce4e3c7e4235%2Fscore_0.png?generation=1754638731130078&alt=media)

**Second stage:**

Otherwise, an MAPE is calculated with the samples that have absolute percentage errors less than or equal to 1. Then the MAPE is divided by the fraction of absolute percentage errors less than or equal to 1. Finally, the score is 1 minus the scaled MAPE.

![score_1.png](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F28096579%2Fca4b376cf10f63393146e5f6d9ced5cc%2Fscore_1.png?generation=1754638744628468&alt=media)

## Submission File
The submission file must contain exactly these two columns: `id` and `new_house_transaction_amount`. Each id is formatted as a month in the form of `%Y %b` and a sector id in the form of `sector n` joined by an underscore. For each id in the test set, you must predict the total new house transaction amount in ten thousand Chinese Yuan for the specific month and sector. The total new house transaction amount is equal to total new house transaction area times transaction price per area. The file should contain a header and have the following format:

    id,new_house_transaction_amount
    2024 Aug_sector 1,49921.00868
    2024 Aug_sector 2,92356.22985
    2024 Aug_sector 3,480269.7819
    etc.

### Submission Requirements

- **Preserve row order**: Maintain the exact same row order as in test.csv
- **Exact Columns**: The submission file should contain the exact columns as in test.csv