Upload README.md with huggingface_hub

README.md

@@ -1,6 +1,99 @@
-# LargeST-GLA Dataset
-Preprocessed from LargeST benchmark (California traffic data).
-Region: GLA
-Year: 2019
-Sensors: 3834
-Compatible with METR-LA and PEMS-BAY format.
+---
+license: mit
+task_categories:
+- time-series-forecasting
+- tabular-regression
+tags:
+- traffic-prediction
+- time-series
+- graph-neural-networks
+- transportation
+- spatiotemporal
+size_categories:
+- 10M<n<100M
+---
+
+# LargeST-GLA Traffic Dataset
+
+## Dataset Description
+
+This dataset contains traffic flow data from the Greater Los Angeles (GLA) area, preprocessed from the LargeST benchmark. It covers 3,834 traffic sensors across 5 counties (Los Angeles, Orange, San Bernardino, Riverside, Ventura) for the year 2019.
+
+## Dataset Structure
+
+### Data Format
+
+- **Format**: Parquet files for efficient loading and analysis
+- **Splits**: train (70%), validation (10%), test (20%) - **temporal splits** preserving chronological order
+- **Features**: Time series traffic flow data with temporal and spatial dimensions
+
+### Split Strategy
+
+- **Temporal splitting**: Data is split chronologically to prevent data leakage
+- **All sensors included**: Each split contains data for all sensors at each time step
+- **Training period**: Earliest 70% of time samples across all sensors
+- **Validation period**: Next 10% of time samples across all sensors
+- **Test period**: Latest 20% of time samples across all sensors
+- **Graph structure preserved**: Spatial relationships maintained in all splits
+
+### Data Schema
+
+- `node_id`: Sensor/node identifier (0-3833 for LargeST-GLA)
+- `t0_timestamp`: ISO timestamp for the prediction target time
+- `x_t*_d0`: Input features at different time offsets
+  - Traffic flow values at 12 historical time steps (t-11 to t+0)
+- `y_t*_d0`: Target values at future time steps
+  - Traffic flow predictions for next 12 time steps (t+1 to t+12)
+
+### Dataset Statistics
+
+- **Region**: Greater Los Angeles Area (GLA)
+- **Time period**: 2019 (full year)
+- **Total sensors**: 3,834 sensors across 5 counties
+- **Counties**: Los Angeles, Orange, San Bernardino, Riverside, Ventura
+- **Temporal resolution**: 5-minute intervals
+- **Prediction horizon**: 1 hour (12 time steps)
+- **Total samples**: ~403 million samples
+
+## Usage
+
+```python
+import pandas as pd
+from datasets import load_dataset
+
+# Load from HuggingFace Hub
+dataset = load_dataset("emelle/LargeST-GLA")
+
+# Or load directly from parquet
+train_df = pd.read_parquet("hf://datasets/emelle/LargeST-GLA/train.parquet")
+val_df = pd.read_parquet("hf://datasets/emelle/LargeST-GLA/val.parquet")
+test_df = pd.read_parquet("hf://datasets/emelle/LargeST-GLA/test.parquet")
+
+print(f"Train records: {len(train_df):,}")
+print(f"Val records: {len(val_df):,}")
+print(f"Test records: {len(test_df):,}")
+```
+
+## Citation
+
+If you use this dataset, please cite the original LargeST paper:
+
+```bibtex
+@inproceedings{liu2023largest,
+  title={LargeST: A Benchmark Dataset for Large-Scale Traffic Forecasting},
+  author={Liu, Xu and Xia, Yutong and Liang, Yuxuan and Hu, Junfeng and Wang, Yiwei and Bai, Lei and Huang, Chao and Liu, Zhenguang and Hooi, Bryan and Zimmermann, Roger},
+  booktitle={Thirty-seventh Conference on Neural Information Processing Systems Datasets and Benchmarks Track},
+  year={2023}
+}
+```
+
+## Original Data Source
+
+This dataset is derived from the LargeST benchmark, preprocessed to be compatible with METR-LA and PEMS-BAY formats for spatiotemporal graph neural network research.
+
+- **Original dataset**: [LargeST on Kaggle](https://www.kaggle.com/datasets/liuxu77/largest)
+- **Compatible with**: METR-LA, PEMS-BAY dataset formats
+
+## License
+
+MIT License