Research map / Architectures and efficiency / RNNs and normalization
AirDelhi: Fine-Grained Spatio-Temporal Particulate Matter Dataset From Delhi For ML based Modeling
Sachin Chauhan, Zeel Bharatkumar Patel, Sayan Ranu, Rijurekha Sen, Nipun Batra
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| normalize time separately for each day instead of across days (ablation) | Random Forest · Delhi (Day) forecasting | test RMSE: 110.49 -> 89.54 | Improved |
| normalize time separately for each day instead of across days (ablation) | XGBoost · Delhi (Day) forecasting | test RMSE: 102.68 -> 84.15 | Improved |
| normalize time separately for each day instead of across days (ablation) | NSGP · Delhi (Day) forecasting | test RMSE: 95.83 -> 86.34 | Improved |
| normalize time separately for each day instead of across days (ablation) | IDW · Delhi (Day) forecasting | test RMSE: 86.52 -> 270.73 | Hurt |
| use learned ML models instead of the Mean Predictor | Random Forest · Delhi (Day) interpolation | RMSE: 65.80 -> 22.24 | Improved |
| use distance weighting (IDW) instead of Mean Predictor | IDW · Delhi (Day) interpolation | RMSE improvement: 39.3% | Improved |
| drop training-day data and use only test-day visible set (ACT to C) (ablation) | XGBoost · Delhi (Day) interpolation | RMSE: 33.24 -> 29.73 | Improved |
| drop training-day data and use only test-day visible set (ACT to C) (ablation) | Random Forest · Delhi (Day) interpolation | RMSE: 22.24 -> 22.24 | No clear effect |
| drop historical training days (ablation) | NSGP · Delhi (Day) interpolation | RMSE | Hurt |
| use recent complex models (NSGP, GraphSage, N-BEATS) instead of Random Forest/XGBoost | NSGP, GraphSage, N-BEATS · Delhi (Day) | RMSE | No clear effect |
| switch task from interpolation to forecasting | best model · Delhi (Day) | RMSE: 22.24 -> 84.15 | Hurt |
| evaluate on Delhi data versus Canada/USA data | ML models vs Mean Predictor · PM2.5 datasets | RMSE improvement over Mean Predictor: Canada ~27% best, USA 16% - 26%, Delhi up to 66.2% | Mixed |
Closest papers by meaning
- Indoor Air Quality Dataset with Activities of Daily Living in Low to Middle-income Communities (NeurIPS 2024 · similarity 0.86)
- AirPhyNet: Harnessing Physics-Guided Neural Networks for Air Quality Prediction (ICLR 2024 · similarity 0.84)
- Air Quality Prediction with Physics-Guided Dual Neural ODEs in Open Systems (ICLR 2025 · similarity 0.82)
- AirRadar: Inferring Nationwide Air Quality in China with Deep Neural Networks (AAAI 2025 · similarity 0.82)
- Measuring Fine-Grained Urban Air Temperature with Satellite Imagery (AAAI 2025 · similarity 0.81)
- UrbanDataLayer: A Unified Data Pipeline for Urban Science (NeurIPS 2024 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
