# DLA4A Test — Data

来源：https://www.kaggle.com/competitions/dla-4-a-test/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

The data provided for this challenge is a subset of the same one used by Kratzert et al. to show that an LSTM approach significantly outperforms the NOAA model.

**File descriptions**
- **DLA_Project-A_challenge_train.csv** - 20 years of data for 100 of the CAMELS U.S. basins; use this dataset for both training and validation, partitioning it however you think is best (e.g. to avoid overfitting).
- **DLA_Project-A_challenge_test.csv** - 10 years of data for the same 100 CAMELS U.S. basins provided in the training and validation set.
- **DLA_Project-A_challenge_test_unseen_basins.csv** - Data for 10 U.S. basins not included in the training and validation set; to be used for more fully evaluating model generalization performance.